• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    The present invention relates to a high-speed imaging sensor system in which single-photon detectors are provided in an architecture adapted for high-speed processing of the output of the detectors with high fidelity to filter out false positives.
    公开号:BE1028366B1
    申请号:E20205564
    申请日:2020-08-07
    公开日:2022-01-11
    发明作者:Der Tempel Ward Van；Johannes Willem Peeters；André Bernard Miodezky
    申请人:Voxelsensors Srl；
    IPC主号:
    专利说明:

    t BE2020/5564 PIXEL SENSOR SYSTEM
    TECHNICAL DOMAIN The invention relates to an improved system and method for high-speed imaging, in which an environment can be reliably captured in limited time spans and strongly limited photon budgets.
    BACKGROUND ART In the state of the art, in scanning active imaging, a light beam, typically a laser, is moved over an area to be captured, and the location where the beam impinges is recorded at each point in time via multiple image sensors. By processing the differences in location from the different points of view (sensors), the effective distance to the illuminated target can be determined, via triangulation. Such a measurement captures a voxel. The speed at which this process can be carried out, the voxel rate, is limited on the one hand by the speed at which the scanning with the light beam takes place, but on the other hand is also (strongly) limited by the processing time required by the sensors to detect the reflected light beams. , especially in relation to background radiation (ambient light) and general thermal noise. By specifically tackling this second issue, imaging can be significantly accelerated. In order to achieve a voxel rate of tens or even hundreds of millions of voxels per second, each voxel must be captured in a time span of at most 10 ns. As a result, the sensor must also be suitable for operating with limited photon budgets (i.e. detected photons incident on the sensor that are sufficient for a detection), such as for instance 10 photons. Given the limited time span for processing, only a limited number of photons can be captured. Existing image processing and shaping systems process optical input acquired by the sensors either in parallel for all pixels, in the case of a so-called global shutter, or staggered over time in the case of rolling shutter. In both cases, typical imaging systems have a gain of 10 UV to 1 mV per incident electron, in order to achieve a signal that exceeds a minimum detection voltage and can be recorded. In the above range, it should be noted that the upper limit of this can only be guaranteed by more recent imagers, which have been specifically modified to count photons, and focus on very low detection rates. With such specialized sensors, a signal of 10 mV can be produced for 10 electrons, which can be read positively for a particular pixel. In order to detect an incident photon packet, a minimum number of 10 photons is required, which moreover therefore impinge on the sensor in a time span of 10 ns. What has been done in the prior art so far is to limit the exposure time of the sensor, for example to 10 ns, and then read the sensor to detect an event via a threshold voltage (ie true incidence of reflected beam instead of a false positive from ambient light or thermal noise). The disadvantage of this is that, especially with high-resolution imagers, the time required for reading the sensor dominates the process, and is typically significantly higher than 10 ns, causing a bottleneck here.
    An alternative is to let the sensor decide for itself whether a minimum amount of photons has been detected within a certain time span, which amounts to the detection of a voxel, instead of reading the sensor and deciding based on the data read whether or not there is a voxel. no event took place. However, a problem with this is that the time it takes for the sensor to evaluate whether or not an event has occurred is still too high, and currently in the most recent versions, such as with the Prophesee sensors, still at least 1 us. Moreover, in such performances it is almost impossible to distinguish false positives, due to ambient light or thermal noise, from real events.
    WO 2013/018006, US 2012/257789 and US 2018/262705 describe related systems in the prior art, which, however, fail to sufficiently solve the discussed problems.
    Currently, there are no systems of sensor architecture capable of achieving the desired detection rates under the proposed characteristics such as photon budget.
    The present invention aims to find a solution to at least some of the above-mentioned problems. SUMMARY OF THE INVENTION
    The invention relates to a high-speed sensor system according to the claims. The current architecture concerns a system in which highly sensitive sensors, such as SPADs (single photon avalanche diodes) are used to detect events at very low photon budgets, which allows a very short scanning time, whether or not real detections or incorrect perception by ambient light or thermal noise. The combination of these features makes it possible to perform imaging very quickly, even at high resolutions.
    DESCRIPTION OF THE FIGURES Figure 1 shows again a schematic representation of a rudimentary version of the system according to the invention. Figure 2 shows a schematic representation of a system (1) according to an embodiment of the invention. Figure 3 shows a schematic representation of a photosensitive zone provided with detectors and connected to row and column buses.
    Figures 4A-C show further embodiments based on Figure 3.
    DETAILED DESCRIPTION Unless otherwise defined, all terms used in the description of the invention, including technical and scientific terms, have the meaning as generally understood by those skilled in the art of the invention. For a better assessment of the description of the invention, the following terms are explicitly explained.
    “A”, “the” and “the” refer to both the singular and the plural in this document unless the context clearly dictates otherwise. For example, “a segment” means one or more than one segment.
    When "about" or "around" in this document is used with a measurable quantity, a parameter, a duration or moment, etc., variations of +/-20% or less, preferably +/-10% or less, more preferably +/-5% or less, even more preferably +/-1% or less, and even more preferably +/-0.1% or less than and of the quoted value, to the extent that such variations of are applicable in the described invention.
    However, this should be understood to mean that the value of the quantity at which the term "approximately"
    or “around” is used, itself is specifically disclosed.
    The terms “comprise”, “comprising”, “consist of”, “consisting of”, “include”, “contain”, “containing”, “include”, “include”, “contain”, “include” are synonyms and are inclusive or open terms designating the presence of the following, and which do not exclude or preclude the presence of other components, features, elements, members, steps known from or described in the art.
    Citing numerical intervals through the endpoints includes all integers, fractions and/or real numbers between the endpoints, including these endpoints.
    In a first aspect, the invention relates to a high-speed imaging sensor system, the system comprising one or more light sources, and comprising an array having a plurality of single-photon detectors (single-photon detector or SPD), which detectors are spatially distributed over the array in an array. substantial matrix form, preferably at regular intervals, the SPD being capable of detecting single photons, and registering the detection of the single photon with a detection signal; the system further comprising a plurality of row buses and column buses, and wherein the detectors are grouped in rows and/or columns, and wherein the detectors per row are connected to one or more row buses and/or wherein the detectors per column are connected to one or more row buses column buses are connected to the column bus for aggregating signals from the detectors, whereby per row bus only detectors from one of the rows are connected to the row bus for aggregating signals from the detectors, whereby per column bus only detectors from one of the columns are connected are; the system further comprising an evaluation circuit, wherein the row buses and the column buses are connected to the evaluation circuit, the evaluation circuit being adapted to evaluate the aggregated signals of the row buses and the column buses according to predetermined confirmation patterns to confirm the detection of a incident photon and localization thereof, said confirmation patterns comprising temporal and spatial conditions, and wherein a detection is confirmed on the basis of a satisfaction of the signals from
    > BE2020/5564 row buses and column buses on a predetermined pattern.
    In addition, the spatial conditions relate to detecting an aggregated signal comprising a detection signal from at least two column buses associated with adjacent columns in a predetermined time window, and detecting an aggregated signal comprising a detection signal from at least two row buses associated with adjacent rows in a predetermined time window. time window.
    Until now, single-photon detectors have had very limited use due to their high sensitivity, which means that false positives have too much impact on the imaging.
    In addition, to remove these false positives, further processing of the output of the SPDs is required, and the processing speed is reduced and the benefits of SPDs are de facto negated.
    The present invention succeeds in filtering out false-positive detection signals by, on the one hand, limiting the input to be processed by grouping the sensors and aggregating the individual signals (row buses and column buses) and processing in this way, using predefined confirmation patterns false events are removed.
    In this way, the (relative) location of the detecting sensor can also be determined again from the aggregated info.
    The evaluation circuit thereby checks the signals from the buses (row and column) and matches them to the confirmation patterns to remove false positives.
    These confirmation patterns may include spatial and/or temporal conditions, spatial conditions being related to the (relative) position of the sensors (and/or their column and/or row buses) whose signal indicates a detection.
    For example, when a signal indicating a detection is received from two column buses whose sensors are adjacent (for example, columns 17 and 18), this can be interpreted as confirmation that it is not a random misfire due to thermal noise, for example.
    Whether or not combined with this, temporal conditions can also be imposed that are related to the (absolute and/or relative) time at which signals indicate a detection, such as the requirement that only on receipt of two (or more) signals indicating a detection within a predetermined time span, this is interpreted as a real event and detection, and is thus confirmed.
    The Applicant noted that in most prior art systems, either the evaluation circuitry received too much input to process and thus could no longer perform it in the planned 10 ns time frame, or the sensors themselves were modified to allow the data to preprocess, resulting in solutions that are particularly impractical and/or expensive, as more sophisticated sensors or add-ons are required to enable this functionality. Moreover, it is not possible for the sensors themselves to reliably process the signals without taking into account the input from other sensors. To solve this, all sensors would have to be interconnected in terms of input, which as mentioned has a high impact on cost price as well as energy efficiency, speed and compactness. In a preferred embodiment, the spatial conditions comprise detecting an aggregated signal comprising a detection signal from at least two column buses associated with adjacent columns in a predetermined time window, and detecting an aggregated signal comprising a detection signal in at least two row buses associated with adjacent rows in a predetermined time window.
    The use of spatial conditions can be applied in the form of so-called coincidence detection, in which the (relative and/or absolute) physical location of the origin of several 'positive' signals is compared in order to be able to confirm these on the basis of probabilistic estimates. For example, in certain circumstances it can be expected that neighboring or adjacent column groups or row groups (possibly with an additional limitation in terms of 'length', the number of sensors per group) have a certain chance of an (essential) simultaneous (in the same period of time) positive signal that is negligible. , for example in conditions with limited ambient light. In other circumstances, the conditions can be made stricter to avoid false positives, for example by requiring three or more adjacent groups to speak of a confirmed detection. Additionally or alternatively, requirements can also be made about the strength of the signals (threshold value) of the buses in order to speak of a positive detection, which further depends on the type of sensor and sensitivity, as well as variation in the gain of the sensor.
    Preferably, use is made of the relative location of the detectors associated with the column buses and row buses in order to verify the spatial conditions. Typically, an array is regularly subdivided, with a fixed number of detectors per column bus and a fixed number of detectors per row bus (whether or not equal for row buses and column buses), which greatly simplifies logic programming for verifying spatial conditions. Once a detection signal has been confirmed, the effective location can then be easily determined from the known position of the detectors of row and column buses. Aggregating the signals from several detectors into one signal for a column or row bus greatly alleviates the computational requirements. The disadvantage that is normally linked to this is that some of the information is lost, namely that detection can only be confirmed for a group of detectors linked to the column or row bus, but it cannot be detected for which detector(s) it is specific. However, by doing this for both row and column buses, a detect can be located along the column buses (equivalent to an X coordinate in the array) and a detect located along the row buses (equivalent to a Y coordinate in the array ), and together giving an approximate 2D position for the location of the detection.
    In a preferred embodiment, the temporal conditions comprise a temporal overlap of sensing an aggregated signal comprising a detection signal of at least one row bus, and preferably at least two row buses associated with adjacent rows, and of sensing an aggregated signal comprising a detection signal of at least one, and preferably at least two column buses associated with adjacent columns, the localization of the incidence of the photon on the array being determined on the basis of the column buses and row buses associated with a confirmed detection.
    Again, the temporal conditions can be set more strict or looser depending on the circumstances, and in certain situations even dynamically in such a way that they automatically adjust based on, for example, average detection of (false-positive or not) signals in the buses per unit of time.
    In general, the temporal conditions require that, in order to confirm a detection, the separate detection signals from row bus(es) and column bus(es) overlap, meaning here that the beginning (leading edge) of the detection signals fall within a predetermined time span of each other, less than 100 ns, such as, for example, less than 50 ns, or preferably even less than 10 ns, 5 ns, 2.5 ns, 2 ns, 1.5 ns, 1.0 ns, or even less. Preferably, the detection signals overlap to a substantial extent (at least 25%, preferably at least 50% or even 75% overlapping). This requirement ensures that random signals due to thermal noise and the like are statistically unlikely as they should be detected almost simultaneously in two different detectors. To increase this reliability, the predetermined time span can be narrowed and/or the degree of overlap, and/or any other factors. Based on this, it can also be verified whether these are separate events when multiple row buses and/or column buses are detected. Finally, upon detection on multiple row buses and multiple column buses, the correct matching can also be performed between the different column buses and row buses by comparing the degree of overlap. If e.g. at t = 0.075 ns a signal is detected on column bus 208, and at t = 2.45 ns on column bus 472; and at t = 2.38 ns on row bus 171 and at t = 0.081 ns on row bus 23, it can be deduced that most likely two events have been detected, one at position (row bus / column bus) 208 / 23 and one at position 472 / 171.
    A second temporal correlation is applied between successive time windows. After all, with a scanning system it is to be expected that, if for instance a correct detection has taken place in time window 1, with detected position, for instance [10,100], the expected detection coordinates for time window 2 will be close to this. In this way, consideration can be made to consider detections that are far from the previously detected coordinate and/or confirm as true or false with possible future information. In this way, a spatio-temporal pattern can be found in the detections over a number of time windows that should match or closely resemble the expected pattern as a function of the scanning pattern of the light source. The concept of the invention thus reduces the computational load, by aggregating signals from detectors, allowing effective events to be detected quickly for a given bus, and then knows by ingeniously using additional constraints to retrieve the precise location of the event. . Particularly preferably, the spatial conditions comprise sensing an aggregated signal comprising a detection signal from at least two column buses associated with adjacent columns in a predetermined time window, and sensing an aggregated signal comprising a detection signal from at least two row buses associated with adjacent rows in a predetermined time window, and wherein said confirmation patterns comprise temporal and spatial conditions.
    By imposing both spatial and temporal conditions, the reliability of confirmation of detection is greatly improved. Moreover, in this way it is also possible to correctly interpret multiple events in a time span as discussed in previous paragraphs.
    In a preferred embodiment, each SPD is uniquely associated with one column bus and uniquely associated with one row bus. The advantage here is that the number of detectors is limited and the location can be determined more accurately (X and Y coordinate). For example, in a row with 20 SPDs associated therewith, the first SPDs can be connected to a first row bus for that row, the second 5 to a second row bus for that row, and the next sets of 5 to a third and fourth row bus for that row. that 5 row.
    Each SPD is therefore only connected to one bus, but conversely, the (driving) buses can be connected to several SPDs.
    In an alternative preferred embodiment, each SPD is uniquely associated with either one column bus or one row bus.
    Such a system will be simpler in architecture, but will use more detectors.
    A strong advantage, however, is that 'false' detections on a detector (by e.g. ambient light and the like) will typically not be passed to both column and row bus associated with the detector, as in the previous embodiment, although such false positives via other conditions can be filtered out (such as requirement that 2 or more adjacent buses must receive a detection signal). In a further preferred embodiment, the array comprises M on N single-photon detectors, which are positioned in a matrix with M rows and N columns, wherein per row the SPDs are distributed over two or more row groups of, preferably in the row consecutive, SPDs, and wherein the same distribution of the SPDs per row continues across all rows, the SPDs of a row group are connected to row bus associated with the row group, and wherein the spatial conditions additionally involve sensing an aggregated signal comprising a detection signal in at least two row buses associated with adjacent row groups in a predetermined time window.
    For example, there are a maximum of M row buses and N column buses, but in practice much less, since several SPDs are grouped per row and column bus.
    In certain circumstances, at very high resolutions, the number of detectors per row is very high, and it is computationally more efficient to further divide the detectors per row into a number of row groups per row (preferably where each detector can belong to only one row group, although options exist in which some or even all detectors belong to two or more row groups), each row group having its own associated row bus.
    Preferably, each bus manages between 1 and 20 detectors, more preferably between 2 and 10, even more preferably between 2 and 5, or even between 2 and 4. This limits the amount of multiple events on a row bus per time span, allowing it to differentiation requires less computing power.
    In addition, with a very large number of detectors per row without subdivision into row groups, the chance increases that several detectors detect an event quasi-simultaneously and cannot differentiate it afterwards, or with limited certainty, when aggregating the signals to the row bus. fall. A suitable choice of number of detectors per row group will greatly limit this problem, and on the basis of statistical observations an optimized number can be determined, which may also depend on the application of the array (much or little ambient light). In certain embodiments, the number of detectors per row group can also be dynamically or manually adjusted to accommodate the circumstances. Thus, there may be a number of predetermined distributions between which to choose. In a further preferred embodiment, the array comprises M on N single-photon detectors which are positioned in a matrix with M rows and N columns, wherein per column the SPDs are distributed over two or more column groups of SPDs, preferably consecutive in the column. , and wherein the same distribution of the SPDs per column continues over all columns, wherein the SPDs of a column group are connected to column bus associated with the column group, and wherein the spatial conditions additionally concern the detection of an aggregated signal comprising a detection signal in at least two column buses associated with adjacent column groups in a predetermined time window. The same reasoning as described earlier for the division of the detectors in a row into row groups applies here. Particularly preferably, both distributions (row groups and column groups) are combined as described above. Alternatively, the detectors can be alternately assigned to the multiple row buses and/or column buses, which offers advantages in the coincidence determination. For example, with an impact that triggers several adjacent detectors in one row (which often happens), this will lead to two or more row buses belonging to the row that detect a detection. With a sequential distribution, this will lead to only one row bus being triggered (unless it concerns just two detectors located on the transition between two row groups).
    In a preferred embodiment, the SPDs are adapted to provide a signal to the connected row bus and/or column bus, said signal being substantially binary in nature and representing whether or not the SPD detects incident photons.
    In a preferred embodiment, the system is adapted to precharge the row buses and column buses at fixed intervals, the row buses and column buses being adapted to unload upon receipt of an aggregated signal including a detection signal. In a preferred embodiment, the evaluation circuitry is adapted to check spatial consistency, wherein a last localization of the photon incidence on the array is compared with one or more previous localizations, and wherein the evaluation circuitry rejects the last localization upon detection of a spatial discrepancy between the latest localization and the previous localizations above a predetermined, dynamic or non-dynamic, upper limit.
    Due to the above adjustment, on the basis of historical data, namely 'known' incidence position of the reflected beam, a rough filtering can be performed that removes false signals without much computing power being spent on this in further processing.
    In a preferred embodiment, the system includes a synchronization component for synchronizing the signals from the detectors. In order to determine the disparity between the detectors and the light sources, a synchronization component is provided which temporally synchronizes and synchronizes them. This can be achieved, among other things, by physically connecting the elements (detectors and light sources) via cable and providing the elements with a synchronization signal via the cable. Alternatively, the optical signal (signal from the light sources) may per se provide a synchronization moment obtained upon acquisition. This can be done, inter alia, by pulsing the optical signal according to a predetermined pattern that is 'recognised', for example as the start of the scanning, whereby all sensors can use this as a local time 0, and are therefore mutually synchronized, as well as with respect to the light sources. In a preferred embodiment, the system is adapted to only consider signals from a variable subset of the detectors during a portion of an imaging procedure into the evaluation circuitry during imaging and not withhold signals from detectors in the subset, with the system incorporating the detectors into selects the subset based on previous localizations and optionally based on orientation and/or positioning of the one or more light sources, wherein the detectors in the subset comprises a maximum of 25%, preferably a maximum of 10%, of the total number of detectors, and wherein the detectors are grouped together in a subset.
    In practice, most of the detectors on the array will not experience an event per time span.
    By adapting the system to only process signals from a part of the detectors in the evaluation circuit (and thus effectively ignoring a part), the computational burden can be greatly reduced.
    The difficulty lies in the fact that it is not possible to determine with certainty in advance which detectors may be ignored.
    For that reason, an estimate can be made on a statistical basis, and historical data, of possible variation of the incidence position of the reflected beam or beams on the array, relative to the previously detected incidence positions, and preferably also on the basis of the scanning pattern of the light source(s). Taking this into account, a zone of interest can be defined where the reflected beam will impinge with a predetermined statistical probability, not considering detectors outside this zone of interest.
    This zone of interest may be substantially circular, oval, rectangular, or jagged in pattern.
    In this way, a substantial part of the detectors, especially with very large arrays, should not be taken into account during processing.
    The determination of the zone of interest, in certain embodiments, may also take into account detections in several of the foregoing time periods, so as to recognize a pattern therein which can be taken into account in predicting the zone of interest.
    A very large, additional advantage of limiting to a subset of the detectors that are active/charged is that in this way a large number of false positives (due to thermal noise and the like) are also not processed.
    In a preferred embodiment, the system is adapted to activate only a variable subset of detectors during imaging during a portion of an imaging procedure and not deactivate detectors in the subset, the system selecting the detectors in the subset based on previous locations and optionally based on orientation and/or positioning of the one or more light sources, wherein the detectors in the subset comprise a maximum of 25%, preferably a maximum of 10%, of the total number of detectors, and the detectors in a subset are grouped together. In a preferred embodiment, the detectors comprise single-photon avalanche diodes (SPAD), and are preferably SPADs.
    SPADs are semiconductor photodetectors that use a very high reverse voltage, such that impact ionization takes place upon detection, causing an avalanche effect, and thus a high current that grows very quickly. The choice for SPADs is partly based on the speed at which the impact ionization is achieved, and the simple 'resetting' of these detectors. In a preferred embodiment, the detectors are provided with a quenching circuit which normalizes the detection signal from the detectors.
    The presence of a quenching circuit is necessary to reduce the signals (avalanche current), to reset the detectors, as well as to limit the signal caused by the avalanche current. The quenching circuit is preferably active, for example using a so-called discriminator which reduces the reverse voltage, but can alternatively also be passive, such as in the form of a resistor in series with the detector. In a preferred embodiment, a maximum of 100 detectors, preferably a maximum of 50 detectors, are connected per row bus and per column bus.
    As indicated, it is more efficient to limit the number of detectors per bus. In this way, it is easier to distinguish multiple events on a bus from each other, and stricter spatial conditions can be imposed.
    In what follows, a number of specific embodiments will be discussed in practice, as well as with reference to schematic representations in the figures.
    EXAMPLES In a first example, the invention relates to a detection architecture adapted for very fast (i.e. over time spans of 10 ns or less) and low-threshold (i.e.
    at limited photon budgets) photon detection, and translation thereof to a clear digital signal to represent an event (effective detection of reflected light beam as opposed to ambient light or thermal noise).
    Figure 1 shows a schematic representation of a rudimentary version of the system according to the invention, comprising at least one light source (2) and one, preferably N, detectors or sensors (1) that capture a particular scene in the environment. The light source illuminates the scene with a light beam (laser or LED), typically in a preprogrammed pattern and/or in a sequential manner, such as scanning the scene through a point of light (3) in a dynamic pattern (4). The N detectors (1) detect the position of the light point (3) in the scene, for example via triangulation based on the output of the different detectors.
    Since the design of the invention is aimed at processing millions of voxels per second, without brute forcing the problem, the scanning speed must be very high, and its processing optimized given the limited photon budget and short time frame. .
    Figure 2 shows a schematic representation of a system (1) according to an embodiment of the invention, the system comprising a photosensitive zone (11), typically comprising an array of photosensitive elements or detectors (e.g. SPAD). The system further includes an evaluation circuit (12), which includes a logic circuit, such as a logic gate or variations thereof, connected to the photosensitive zone (11) and adapted to determine the validity of the data. obtained from the detectors of the photosensitive zone. Furthermore, the system here comprises a logic circuit (12 when combined with logic circuit for validating the data) for merging the data/signals obtained from the detectors into a (single) data stream. The system may also include a synchronization component for synchronizing the data stream and the optical signal from the light source(s) and/or the data from the different detectors.
    In a first more specific version of this example, conceptually depicted in Figure 3, the photosensitive zone (11) comprises a plurality of SPADs (21) arranged in an array of M rows and N columns, each SPAD having a quenching circuit, passive or active, and wherein the SPADs are represented as a square cell. Note that the two connection points per cell are purely visual, serving as the connection point from the SPAD to a row and/or column bus. All SPADs are connected to a column bus (24) and to a row bus (23), whereby one or more buses are provided per row and/or column (preferably several buses per row and per column, although it is possible to have all SPADs on a row or column on one bus). In Figure 3, 2 row buses are used per row and 2 column buses per column, whereby the SPADs are connected alternately to one or the other bus. Alternatively, this can also be done sequentially (first X consecutive SPADs on first bus, rest on second bus). Detection of a photon on SPAD will lead to detection of an event on the associated row and column bus. Figure 4A shows the example of a photon impact detected on 4 different, adjacent SPADs (see circle). If each impact is sufficient to trigger a detection on the SPAD (low energy impact may be insufficient to cause the SPAD to cause avalanche breakdown), a detection is propagated on 4 row (group) buses and 4 column (group) buses, as shown in can be seen by means of the signals on the left for the row buses, and below for the column buses.
    Also in Figure 4B, an event is detected on a single SPAD (top left, second row, first column), leading to a propagated detection on one row bus and one column bus.
    The evaluation circuit (12) which receives the signals from the row and column buses hereby imposes a number of conditions for evaluating the incoming signals, to indicate whether or not a confirmed event. For example, the evaluation circuitry will typically regard solitary events, such as the top left detection, as a false detection, while multiple detections on neighboring buses (such as on row 3-4, column 3-4) will typically be considered true events.
    In Figure 4C, an additional solitaire event is detected on row 5, column
    2. This leads to a supposed event on two adjacent column buses (column bus 2 and 3), which, however, originates from two separate events. However, by also taking the busses into account, it can be seen that these are separate solitaire events, both of which can be retained.
    To avoid that solitary events can lead to a pattern that satisfies the conditions of the evaluation circuit (predetermined pattern or patterns), it is advantageous to provide enough buses per row and column of SPADs, as this limits the chance of hitting neighboring SPADs. buses happen to perceive events in the same time frame.
    However, real events (ie real impact of reflected light beam on array) will normally lead to detections that meet the conditions of the evaluation circuit, as the triggering is not limited to one SPAD, but will also trigger one or more neighboring SPADs, and moreover, in a very limited time frame, which will confirm the correlation between the individual detections when evaluated.
    The invention hereby makes use of coincidence detection based on the spatial position of the pixels (detectors) in the array, by looking for confirmation patterns in the signals of the row and column buses, in order to determine whether a real photon packet (and thus a reflected light beam) was detected. In Figure 4C, an additional measure is provided, namely limiting the active/charged detectors to a subset of the detectors, by means of a so-called “enable window” (22). The term “enable window” carries no limitation on shape or size, and may vary dynamically herein. This window acts as a predictive zone in which the light point is expected to be detected based on historical data (such as previous detections, movement pattern over previous detections, ambient light strength, etc.). In this way, a substantial part of the SPADs are not taken into account for the localization of incidence of the light beam, which on the one hand greatly reduces the false positives (since they are largely randomly distributed, and by limiting the number of SPADs , the number of false detections is limited to the same degree), and on the other hand also limits the number of signals to be processed, and the buses are prevented from being saturated by solitary events. Alternatively, or in addition to the foregoing, the sensitivity of the detectors can also be adjusted, dynamically or not, in order in this way to avoid saturation of the column and/or row buses. As a result, more energy will have to be used in the active optical signal (i.e. light source) to increase the chance of sufficient photons coincident (coincidence). In this way background radiation and thermal noise can be filtered efficiently. By defining a number of confirmation patterns, the system can be easily, quickly (and possibly dynamically) adapted for different situations. For example, in situations with limited ambient light, such as at night, indoor, etc., the spatial condition can be removed or relaxed. Also, 'unwanted' events can be filtered out by comparing the raw detections to an expected lighting pattern, which takes into account, for example, the pattern with which the lighting source scans the scene. Thus, in one embodiment, the light source may be programmed to scan the scene according to a Lissajous pattern, which will impact the sensed detections, which will typically adopt a similar pattern, albeit distorted. In this case, strongly deviating detections from the expected illumination pattern can be rejected as noise and the like.
    In addition, in certain embodiments, multiple "confirmation streams" of evaluated detections matching different confirmation patterns may be held in parallel. In this way, these can be checked against measurements in later time periods in order to achieve greater certainty which the correct confirmation flow was.
    The communication from detector to bus can, for example, be done via a pull-down (or pull-up) element on the bus, whereby a detection generates a signal that causes discharge of the precharged bus. At the end of each time span of detection, the buses are preloaded again to wait for the occurrence or not of a detection from the SPAD. The advantage of this is that these pull-down discharges are very easily detectable by the evaluation circuit. Optionally, an amplifier circuit may be provided prior to discretizing the signals. Based on this, the system can provide a list per time span of locations in the array where confirmation patterns were detected.
    In an alternative embodiment, the invention relates to a system in which the photosensitive part is implemented on 1 layer and the processing circuitry and logic on a 2nd layer. The layers are stacked on top of each other and connected with 1 or more electrical connections per pixel or per pixel group. The search for expected detection patterns in the sensor can now be done even more locally, whereby instead of detection patterns on projections (row/column buses), a two-dimensional pattern can be searched in each sub-window of the sensor array. Another embodiment of the invention introduced a stacking layer in which the spad detector triggers are combined and analysed. When working on a single layer, the assertions need to be monitored on the 'projections', either on column and row busses, or different projections such as diagonal or other busses. The introduction of a second layer provide more local monitoring of spad detector clusters, monitoring their co-inciding event creation. Due to the more local monitoring, less spad detector devices can be considered per cluster and saturation of the busses can be avoided when high amounts of ambient event are being generated.
    Also, the assertion pattern can now be considered as a 2 dimensional assertion pattern that can be compared with each detection cluster under test. When a match is found with the wanted assertion pattern a detection can be considered with high confidence.
    In an alternative embodiment, the invention relates to an array of pixels, wherein each pixel comprises one or more single-photon detectors, preferably SPADs. The pixels can (partially) overlap, whereby neighboring pixels can share detectors.
    Due to thermal noise and background radiation incident on the pixels, the detectors will emit a detection signal at a rate equal to the DCR (dark count rate) and BGR (background rate). These events are not temporally correlated with each other, so the detectors will be triggered separately. By including in the confirmation pattern or temporal and spatial conditions that a detection is only confirmed when one or more SPADs in a pixel are triggered in a predetermined time frame, the acceptance of false positives is drastically reduced, since false positive events are not be correlated. In contrast, when a pixel is irradiated with an active light beam (i.e. from illumination of the scene by the light source(s)) within the predetermined time span (of e.g. 10 ns), the energy budget of the light beam is adjusted such that a SPAD or other detector illuminated by a reflected beam also effectively detects and triggers one or more photons. In this way, the probability also greatly increases that a pixel detects multiple events in the individual detectors or SPADs of the pixel. This further relies on coincidence detection. In the present invention, the principle of coincidence detection is used as a signal without providing an exact timestamp associated with the event, since only the location of the event on the array is needed. In this way, a pixel can detect coincidence events and pass it on to the column and row bus to indicate that an event has occurred. The coincidence evaluation mechanism may be provided per pixel or alternatively per set of pixels. Alternatively, it can also be integrated in the periphery of the systems, which greatly simplifies the array itself, since no local coincidence evaluation is needed anymore, and only quenching circuits. Each SPAD will pass an observation to the row and/or column bus to which the SPAD is associated. The periphery can herein comprise a massive parallel connected digital circuit, which monitors the state of the bus over every predetermined period of time (10 ns). If in this time span events are detected in: - 2 neighboring columns and for the same line (in case a 2X1 confirmation pattern is sufficient); or - 2 adjacent lines and for the same column (for a 1X2 confirmation pattern); or - 2 adjacent lines and two adjacent columns (for a 2x2 confirmation pattern); or - another predetermined confirmation pattern; then an event is confirmed and reported. It can then optionally be time-stamped and communicated to a further device.
    It is believed that the present invention is not limited to the embodiments described above and that some modifications or changes may be added to the described examples without revising the appended claims.
    权利要求:
    Claims (15)
    [1]
    A high speed imaging sensor system, wherein the system comprises one or more light sources, and comprises an array with a plurality of single photon detectors (single photon detector or SPD), which detectors are spatially distributed throughout the array in a substantial matrix form, at preferably at regular intervals from each other, the SPD being capable of detecting single photons, and registering the detection of the single photon with a detection signal; the system further comprising a plurality of row buses and column buses, and wherein the detectors are grouped in rows and/or columns, and wherein the detectors per row are connected to one or more row buses and/or wherein the detectors per column are connected to one or more row buses column buses are connected to the column bus for aggregating signals from the detectors, whereby per row bus only detectors from one of the rows are connected to the row bus for aggregating signals from the detectors, whereby per column bus only detectors from one of the columns are connected are; the system further comprising an evaluation circuit, wherein the row buses and the column buses are connected to the evaluation circuit, the evaluation circuit being adapted to evaluate the aggregated signals of the row buses and the column buses according to predetermined confirmation patterns to confirm the detection of a incident photon and localization thereof, wherein said confirmation patterns include temporal and spatial conditions, and wherein a detection is confirmed by a compliance of the row bus and column bus signals with a predetermined pattern, wherein the spatial condition involves sensing an aggregated signal comprising a detection signal from at least two column buses associated with adjacent columns in a predetermined time window, and comprising detecting an aggregated signal comprising a detection signal in at least two row buses associated with adjacent rows in a predetermined time window.
    [2]
    The high-speed imaging sensor system of claim 1, wherein the temporal conditions involve a temporal overlap of sensing an aggregated signal including a sensing signal from at least one lane bus, and preferably at least two lane buses associated with adjacent rows, and sensing an aggregated signal comprising a detection signal of at least one, and preferably at least two column buses associated with adjacent columns, the localization of the incidence of the photon on the array being determined on the basis of the column buses and row buses associated with a confirmed detection, and preferably according to claim 2 and wherein said confirmation patterns comprise temporal and spatial conditions.
    [3]
    The high speed imaging sensor system of any one of claims 1 to 2, wherein each SPD is uniquely associated with one column bus and uniquely associated with one row bus.
    [4]
    The high-speed imaging sensor system of the preceding claim 3, wherein the array comprises M on N single-photon detectors arranged in a matrix with M rows and N columns, wherein per row the SPDs are distributed over two or more row groups of, at preferably consecutive, SPDs in the row, and wherein the same distribution of the SPDs per row is continued across all rows, wherein the SPDs of a row group are connected to row bus associated with the row group, and where the spatial conditions additionally concern the observation of an aggregated signal comprising a detection signal in at least two row buses associated with adjacent row groups in a predetermined time window.
    [5]
    A high-speed imaging sensor system according to any one of the preceding claims 3 or 4, wherein the array comprises M on N single-photon detectors arranged in a matrix having M rows and N columns, wherein per column the SPDs are distributed over two or more column groups of SPDs, preferably consecutive in the column, and wherein the same distribution of the SPDs per column continues over all columns, wherein the SPDs of a column group are connected to column bus associated with the column group, and where the spatial conditions additionally observe refers to an aggregated signal comprising a detection signal in at least two column buses associated with adjacent column groups in a predetermined time window.
    [6]
    A high speed imaging sensor system according to any one of claims 1 to 5, wherein the SPDs are adapted to provide a signal to the connected row bus and/or column bus,
    wherein said signal is substantially binary in nature and represents whether or not the incident photon is detected by the SPD.
    [7]
    The high speed imaging sensor system of any one of claims 1 to 6, wherein the system is adapted to precharge the row buses and column buses at fixed intervals, the row buses and column buses being adapted to unload upon receipt of an aggregated signal comprising a detection signal.
    [8]
    The high-speed imaging sensor system of any one of claims 1 to 7, wherein the evaluation circuitry is adapted to check spatial consistency, wherein a final localization of the photon incidence on the array is compared with one or more previous localizations and wherein the evaluation circuit rejects the latest location upon detection of a spatial discrepancy between the latest location and the previous locations above a predetermined, dynamic or non-dynamic, upper limit.
    [9]
    The high-speed imaging sensor system of any one of claims 1 to 8, wherein the system includes a synchronization component for synchronizing the signals from the detectors.
    [10]
    A high-speed imaging sensor system according to any one of claims 1 to 9, wherein the system is adapted to consider only signals from a variable subset of the detectors into the evaluation circuitry during a portion of an imaging procedure into the evaluation circuitry during the imaging and signals. of detectors in the subset, where the system selects the detectors in the subset based on previous localizations and optionally based on orientation and/or positioning of the one or more light sources, where the detectors in the subset are up to 25%, at preferably comprises a maximum of 10% of the total number of detectors, and wherein the detectors are grouped together in a subset.
    [11]
    The high speed imaging sensor system of any one of claims 1 to 10, wherein the system is adapted to activate only a variable subset of detectors during imaging during a portion of an imaging procedure and not deactivate detectors in the subset, wherein it system selects the detectors in the subset based on previous localizations and optionally based on orientation and/or positioning of the one or more light sources, whereby the detectors in the subset are a maximum of 25%, preferably a maximum of 10%, of the total number of detectors and wherein the detectors are grouped together in a subset.
    [12]
    A high-speed imaging sensor system according to any one of claims 1 to 11, wherein the detectors comprise single-photon avalanche diodes (SPAD) and are preferably SPADS.
    [13]
    A high-speed imaging sensor system according to any one of claims 1 to 12, wherein the detectors include a quenching circuit that normalizes the detection signal from the detectors.
    [14]
    A high-speed imaging sensor system according to any one of the preceding claims 1 to 13, wherein a maximum of 100 detectors, preferably a maximum of 50 detectors, are connected per row bus and per column bus.
    [15]
    The high-speed imaging sensor system of any one of claims 1 to 14, wherein the evaluation circuitry is adapted to check spatial consistency, wherein a final localization of the photon incidence on the array is compared with one or more previous localizations and wherein the evaluation circuit rejects the latest location upon detection of a spatial discrepancy between the latest location and the previous locations above a predetermined, dynamic or non-dynamic, upper limit.
    类似技术:
    公开号 | 公开日 | 专利标题
    EP0664012B1|1997-05-14|Method and apparatus for classifying movement of objects along a passage
    US8469261B2|2013-06-25|System and method for product identification
    US9177385B2|2015-11-03|Object counter and method for counting objects
    EP2312336A1|2011-04-20|Image processing device
    US20130235364A1|2013-09-12|Time of flight sensor, camera using time of flight sensor, and related method of operation
    JP5826525B2|2015-12-02|Radiation monitoring system and method for detecting radiation source in monitoring area
    CN101569194A|2009-10-28|Network surveillance system
    BE1028366B1|2022-01-11|PIXEL SENSOR SYSTEM
    US20200175238A1|2020-06-04|Reading optical codes
    US9115990B2|2015-08-25|Geospatial and image data collection system including image sensor for capturing 3D geospatial data and 2D image data and related methods
    EP3605148A1|2020-02-05|Solid-state imaging device
    Atluri et al.2009|Development of a sensor system for traffic data collection
    WO2021240455A1|2021-12-02|Pixel sensor system
    US20210325514A1|2021-10-21|Time of flight apparatus and method
    US20200326414A1|2020-10-15|Neuromorphic Single Photon Avalanche Detector | Array Microchip
    JP4085844B2|2008-05-14|Image processing device
    US20220035011A1|2022-02-03|Temporal jitter in a lidar system
    US6069358A|2000-05-30|Pulsed optical threat detection
    KR20210128925A|2021-10-27|Method and system for detection of reflected light pulses in the presence of ambient light
    KR20210073468A|2021-06-18|Nuclear facility unmanned surveillance and artificial intelligence-based automatic alarm system
    US20210302548A1|2021-09-30|Method and apparatus for determining measurement information and lidar device
    WO2021046547A1|2021-03-11|Processing of lidar images
    CN113906313A|2022-01-07|High spatial resolution solid-state image sensor with distributed photomultipliers
    CN114078100A|2022-02-22|Clustering noise reduction device, method, chip, event imaging device and electronic equipment
    CN113412413A|2021-09-17|System and method for imaging and sensing vibrations
    同族专利:
    公开号 | 公开日
    BE1028366A1|2022-01-04|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US20120257789A1|2011-04-06|2012-10-11|Jun Haeng Lee|Method and Apparatus for Motion Recognition|
    WO2013018006A1|2011-08-03|2013-02-07|Koninklijke Philips Electronics N.V.|Position-sensitive readout modes for digital silicon photomultiplier arrays|
    US20180262705A1|2017-03-08|2018-09-13|Samsung Electronics Co., Ltd.|Image processing device configured to regenerate timestamp and electronic device including the same|
    法律状态:
    2022-02-09| FG| Patent granted|Effective date: 20220111 |
    优先权:
    申请号 | 申请日 | 专利标题
    BE202005385|2020-05-29|PCT/IB2021/054688| WO2021240455A1|2020-05-29|2021-05-28|Pixel sensor system|
    [返回顶部]