• 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 invention relates to a device for identifying an index (Clo) of a clothing family, which index is particularly representative of the type and level of clothing of a passenger of a motor vehicle, comprising: at least one camera (2) arranged to capture at least one image of a portion of the passenger clothing (P), - a processing unit (3) arranged to determine at least one index (Clo) of clothing family using a recognition algorithm based on a model of the neural network type, in particular trained on a library (5) of representative images captured by the camera, preferably images previously categorized by a human operator.
    公开号:FR3067835A1
    申请号:FR1755507
    申请日:2017-06-16
    公开日:2018-12-21
    发明作者:Daniel NEVEU；Josselin Gour
    申请人:Valeo Systemes Thermiques SAS；
    IPC主号:
    专利说明:

    The invention relates to a thermal management system for a motor vehicle. The invention also relates to a thermal management process implemented by such a thermal management system.
    In a motor vehicle, it is known to provide management of the flow rates, temperatures and distribution of the air blown by the different aerators as a function of the external temperature and sunshine conditions. On certain vehicles, this can be combined with the activation of a heated steering wheel and / or a heated or cooled seat, and sometimes contact heated surfaces such as an elbow rest.
    The detection and / or taking into account of the thermal state of the passengers is almost nonexistent, except for a few examples of the use of infrared sensors which detect the surface temperature of the passengers' clothes to better take account of the initial conditions during the transitional phase reception (when the person comes from a cold or warm atmosphere) and thermal balance resulting from radiative and convective exchanges. In general, the measurement of the thermal state of the passenger compartment is limited to a measurement of air temperatures combined with a sun sensor.
    More sophisticated approaches to comfort management have been proposed, based on new sensors, in particular infrared cameras, and new actuators, in particular radiant panels and / or localized air supplies.
    The invention aims in particular to propose an improvement of known thermal management systems.
    In particular, this innovation aims to take clothing into account in the management of comfort. It is applicable to the context of a motor vehicle but can also apply to public transport vehicles where it is desired to adapt the parameters for managing thermal comfort to the clothing of each passenger.
    Clothing is known to reduce heat exchanges between the body and the environment and consequently modify the equilibrium temperature of the body and the skin, for a given metabolic activity and environment. This results in a change in thermal sensations and thermal comfort, knowing that comfort is largely correlated with the skin temperature of the different parts of the body.
    Thermo-physiological models that describe a person's thermal comfort, like the Fanger model, take into account the level of clothing in the calculation of heat exchanges. A specific unit is used to characterize the level of clothing of a person: the "Cio", which represents an index of the rate of clothing, which typically varies in a typical range from 0 to 2
    The average person’s clothing, pants with a shirt and jacket for a man, is 1 Cio. A person in a swimsuit will have a clothing index of 0.1 to 0.2 Clos. A person dressed in warm clothing for skiing will have an index close to 2 Clos.
    Knowledge of this clothing index makes it possible to determine the extent of the body's thermal exchanges with the outside atmosphere and to deduce a comfort index. In the Fanger model, we carry out a balance on the whole body by reasoning on an average global temperature.
    The following patent applications can be cited on these aspects: JP2005306095 and US2006144581.
    The subject of the invention is therefore a device for identifying a clothing family index (Cio), an index in particular representative of the type and level of clothing of a passenger of a motor vehicle, comprising:
    - at least one camera arranged to capture at least one image of part of the passenger's clothing,
    a processing unit arranged to determine at least one clothing family index (Cio) using a recognition algorithm based on a neural network type model, in particular trained on a library of representative images captured by the camera, images preferably previously categorized by a human operator.
    The advantage of such an identification method is notably to dispense with the intervention of passengers to describe their clothing and to be able to enrich and improve the robustness of the identification on the basis of feedback. It also makes it possible to take into account a very wide variety of situations and types of clothing by the associative power and classification of a learning by neural network.
    The artificial neural network, as a system capable of learning, implements the principle of induction, that is to say learning by experience. The neural network is generally used in problems of a statistical nature, such as automatic classification.
    According to one aspect of the invention, the camera is a camera operating in the near infrared, in particular in a radiation sensitivity band between 0.5 and 1 μm.
    This strip makes it possible to visualize the passengers day or night, and without visual disturbance for the passengers. This camera is in particular of the NIR (Near Infra Red) type equipped with illuminating lamps.
    As a variant, the camera is an infrared thermal camera, known in English as LWIR (Long Wave Infra-Red), operating in the wavelength band 8 to 13 pm, or a conventional visible camera, or else the images provided by the combination of several cameras, for example an NIR camera and an LWIR camera.
    According to one aspect of the invention, the position, the orientation and the angle of view of the camera (s) are chosen so that at least part of the clothing of the people analyzed is in the field of each camera.
    According to one aspect of the invention, the camera is arranged frontally facing a passenger, for example on the dashboard for the front passengers. In this case, the camera is arranged in such a way that the clothing of the upper body - typically the neck, the skull and if possible the shoulders - is visible.
    According to one aspect of the invention, the camera is arranged in a module, in particular a dome type module, fixed to the ceiling of the passenger compartment. In this case, the camera is arranged to observe several people in its field. The camera is arranged to view part of their clothing, in particular the upper part of the body.
    According to one aspect of the invention, the camera is arranged in a lateral upright of the vehicle (this upright is also called pillar - front or middle pillar). In this case, the camera is arranged to observe several people and to visualize the upper part of the body, in particular the torso and the head, that is to say in the field.
    According to one aspect of the invention, in the case where several cameras are used, they can be placed in the same place to observe a passenger. One can for example have a camera, in particular an NIR camera, facing the driver's face, behind the steering wheel, and a camera, in particular an IR camera, arranged in a ceiling lamp at the level of the ceiling.
    According to one aspect of the invention, the images in the library are classified by clothing family.
    According to one aspect of the invention, the recognition of the level of clothing from the images supplied by the camera or cameras, using a recognition algorithm based on a “neural network” type model which will have been previously trained on a library of representative and categorized images in one or more clothing families. The model is then able to categorize each image received by assigning it one or more clothing families.
    According to one aspect of the invention, a clothing family is representative of a type and level of clothing. The library includes for example four families to describe the clothing, for example in terms of layers on the bust:
    - no layer, for example shirtless: the Cio index is nClol, for example equal to 0.2
    - light layer, for example corresponding to a T-shirt or a shirt or tank top: the Cio index is nClo2, for example equal to 0.5
    - intermediate layer, for example corresponding to a sweater or a jacket: the Cio index is nClo3, for example equal to 1
    - a warm layer, for example corresponding to a coat or a puffer jacket: the Cio index is nClo4, for example equal to 1.5.
    According to one aspect of the invention, the number of families for the bust can be between 2 and 6, and the number of families for the legs can be between 2 and 4.
    According to one aspect of the invention, the families for the legs are for example:
    Shorts
    - Short skirt
    - Long skirt
    - Trousers.
    According to one aspect of the invention, a family can be provided to describe accessories:
    - No scarf
    - A scarf
    - A hat
    - A hat.
    According to one aspect of the invention, each image of the library can be categorized into several families according to the usable information.
    Whatever the quality and richness of the learning phase, it is not statistically possible to have a model with 100% reliability. The invention aims to achieve maximum reliability through a strategy to manage the risk of error and enrich and make the model more reliable.
    According to one aspect of the invention, the processing unit is arranged for:
    - determine a confidence rate on the recognition of a family,
    - when this confidence rate is above a predetermined threshold, for example a 90% threshold, validating the clothing index without passenger intervention,
    - possibly inform the passenger of the family or families selected to manage their comfort.
    According to one aspect of the invention, if the confidence rate is lower than a predetermined threshold, for example a threshold of 50%, the processing unit is arranged to require by default validation of the choice of family by the passenger for activate comfort management accordingly.
    According to one aspect of the invention, if the confidence rate is in an intermediate range, between the two aforementioned thresholds, the processing unit is arranged to allow the passenger the possibility of modifying the family if necessary.
    The categorization or correction of families associated with new images is used, if necessary, to enrich the library of images used to train the model. An update of the model is carried out by a new training when the image library has been significantly modified.
    According to one aspect of the invention, in the case where several cameras are used, different strategies are possible to take advantage of the combination of the different cameras.
    Each camera can be associated with its own recognition model and categorize the images received independently of the other cameras.
    According to one aspect of the invention, the combination of logical and statistical rules can then make it possible to generate a categorization with a highest confidence rate, by combining the categorization and confidence rate of each of the available cameras.
    It is also possible to generate an image library by merging the data from the various cameras and training the model by categorizing each situation from the images of several cameras.
    When the cameras used generate a video stream with, for example, several images per second, it is possible to use images taken at various times to categorize a passenger by exploiting the redundancy of the measurements.
    According to one aspect of the invention, the processing unit can be arranged to operate the model on several images taken at regular intervals, by analyzing the consistency between the different categorizations returned and the associated confidence rates. A statistical approach implemented by the processing unit (occurrences of a categorization with the highest cumulative confidence rate) can make it possible to define the categorization which has the best confidence rate.
    According to one aspect of the invention, an additional step implemented by the processing unit makes it possible to modulate the confidence rate associated with a categorization as a function of the statistical occurrence of a family as a function of other contextual parameters.
    The occurrence of each clothing family can be weighted according to various parameters:
    - The season: low occurrence "warm layer, scarf .." for the summer; low occurrence "light layer" for winter, etc ...
    - The place: low occurrence "naked torso" if no beach or body of water nearby, low occurrence "scarf" if starting in covered parking, etc ...
    According to one aspect of the invention, the categorization is associated in part with a scale of thermal insulation of a part of the body, in particular the bust or the legs, on a progressive scale from 1 to N, N notably greater than or equal to 4.
    According to one aspect of the invention, the categorization adopted takes into account the date, time and place of the shots to weight the confidence rate assigned to each category of clothing according to their statistical occurrence with respect to the parameters. supra.
    According to one aspect of the invention, the processing unit is arranged for:
    - take into account the results obtained on several images to identify the most convincing result by statistical analysis of the occurrences and confidence rate of the results,
    - optionally, take into account other contextual parameters allowing to reinforce the statistical probability of a result, such as the outside temperature or the location of the vehicle.
    The invention also relates to a method for identifying a clothing family index (Cio), an index in particular representative of the type and level of clothing of a passenger of a motor vehicle, comprising the following steps :
    - capture at least one image of part of the passenger's clothing,
    - determine at least one clothing family index (Cio) using a recognition algorithm based on a neural network type model, in particular trained on a library of representative images captured by the camera, preferably images previously categorized by a human operator.
    The invention will be better understood and other details, characteristics and advantages of the invention will appear on reading the following description given by way of nonlimiting example with reference to the appended drawing in which:
    - Figure 1 illustrates, schematically and partially, a device according to the invention,
    FIG. 2 illustrates the exchange of data linked to the device of FIG. 1,
    - Figure 3 shows photos of the library used by the device of Figure 1.
    FIG. 1 shows a device 1 for identifying a clothing family index (Cio), an index representative of the type and level of clothing of a passenger of a motor vehicle V, comprising:
    a plurality of cameras 2 arranged to capture IM images of part of the clothing of a passenger P,
    a processing unit 3 arranged to determine a clothing family index (Cio) using a recognition algorithm based on a neural network type model, linked to a library 5 of representative IM images captured by the camera .
    One of the cameras 2 is a NIR camera operating in the near infrared, in particular in a radiation sensitivity band between 0.5 and 1 μm.
    Another camera 2 is a long wave infrared camera or an English LWIR (Long Wave Infra-Red) camera operating in the 8 to 2 μm band.
    Other types of cameras are of course usable.
    The position, orientation and angle of view of the camera (s) 2 are chosen so that at least part of the clothing of the people P analyzed is in the field of each camera.
    One of the cameras 2 is placed frontally facing a passenger, for example on the dashboard for the front passengers.
    In this case, the camera is arranged in such a way that the clothing of the upper body - typically the neck, the skull and if possible the shoulders - is visible.
    One of the other cameras 2 is arranged in a module, in particular a dome type module, fixed to the ceiling 7 of the passenger compartment. In this case, the camera is arranged to observe several people P in its field. The camera is arranged to view part of their clothing, in particular the upper part of the body.
    One of the other cameras is placed in a side pillar 6 of the vehicle (this pillar is also called pillar - front or middle pillar).
    The cameras 2 can be placed in the same place to observe a passenger. One can for example have a camera, in particular an NIR camera, facing the driver's face, behind the steering wheel, and a camera, in particular an IR camera, arranged in a ceiling lamp at the level of the ceiling.
    The images in library 5 are classified by clothing family.
    As illustrated in FIG. 2, the recognition of the level of clothing from the images provided by the camera (s) 2, using a recognition algorithm based on a model 10 of “neural network” type which will have been previously trained on a library of images 5 representative and categorized into one or more clothing families. The model is then able to categorize each image received by assigning it one or more clothing families.
    A clothing family is representative of a type and level of clothing. Library 5 includes, for example, four families to describe clothing, for example in terms of layers on the bust:
    - no layer, for example shirtless: the Cio index is nClol, for example equal to 0.2
    - light layer, for example corresponding to a T-shirt or a shirt or tank top: the Cio index is nClo2, for example equal to 0.5
    - intermediate layer, for example corresponding to a sweater or a jacket: the Cio index is nClo3, for example equal to 1
    - a warm layer, for example corresponding to a coat or a down jacket: the Cio index is equal to nClo4, for example equal to
    1.5.
    The number of families for the bust can be between 2 and
    6, and the number of families for the legs can be between 2 and 4.
    According to one aspect of the invention, the families for the legs are for example:
    - Shorts
    - Short skirt
    - Long skirt
    - Trousers.
    According to one aspect of the invention, a family can be provided to describe accessories:
    - No scarf
    - A scarf
    - A hat
    - A hat.
    The processing unit 3 is arranged for:
    - determine a confidence rate on the recognition of a family,
    - when this confidence rate is above a predetermined threshold, for example a 90% threshold, validating the clothing index without passenger intervention,
    - possibly inform the passenger of the family or families selected to manage their comfort.
    According to one aspect of the invention, if the confidence rate is lower than a predetermined threshold, for example a threshold of 50%, the processing unit is arranged to require by default validation of the choice of family by the passenger for activate comfort management accordingly.
    According to one aspect of the invention, if the confidence rate is in an intermediate range, between the two aforementioned thresholds, the processing unit is arranged to allow the passenger the possibility of modifying the family if necessary.
    Each camera can be associated with its own recognition model and categorize the images received independently of the other cameras.
    According to one aspect of the invention, the combination of logical and statistical rules can then make it possible to generate a categorization with a highest confidence rate, by combining the categorization and confidence rate of each of the available cameras.
    It is also possible to generate an image library by merging the data coming from the various cameras and training the model by categorizing each situation from the images of several cameras.
    When the cameras used generate a video stream with, for example, several images per second, it is possible to use images taken at various times to categorize a passenger by exploiting the redundancy of the measurements.
    An additional step implemented by the processing unit makes it possible to modulate the confidence rate associated with a categorization as a function of the statistical occurrence of a family as a function of other contextual parameters.
    权利要求:
    Claims (10)
    [1" id="c-fr-0001]
    1. Apparatus for identifying a clothing family index (Cio), an index in particular representative of the type and level of clothing of a passenger of a motor vehicle, comprising:
    - at least one camera (2) arranged to capture at least one image of part of the passenger's clothing (P),
    - a processing unit (3) arranged to determine at least one clothing family index (Cio) using a recognition algorithm based on a neural network type model, in particular trained on a library (5) of images representative captured by the camera, preferably images previously categorized by a human operator.
    [2" id="c-fr-0002]
    2. Device according to the preceding claim, the camera being a camera operating in the near infrared (NIR) in particular in a radiation sensitivity band between 0.5 and 1 pm.
    [3" id="c-fr-0003]
    3. Device according to claim 1, the camera being an infrared thermal camera operating in the wavelength band 8 to 13 pm, or a conventional visible camera, or else the images provided by the combination of several cameras, for example a NIR camera and LWIR camera.
    [4" id="c-fr-0004]
    4. Device according to one of the preceding claims, the camera being arranged frontally facing a passenger, for example on the dashboard for the front passengers. In this case, the camera is arranged in such a way that the clothing of the upper body - typically the neck, the skull and if possible the shoulders - is visible.
    [5" id="c-fr-0005]
    5. Device according to one of the preceding claims, the images (IM) of the library (5) being classified by clothing family.
    [6" id="c-fr-0006]
    6. Device according to one of the preceding claims, the library comprising four families for describing the clothing, for example in terms of layers on the bust:
    - no layer, for example shirtless: the Cio index is nClol, for example equal to 0.2
    - light layer, for example corresponding to a T-shirt or a shirt or tank top: the Cio index is nClo2, for example equal to 0.5
    - intermediate layer, for example corresponding to a sweater or a jacket: the Cio index is nClo3, for example equal to 1
    - a warm layer, for example corresponding to a coat or a down jacket: the Cio index is equal to nClo4, for example equal to
    1.5.
    [7" id="c-fr-0007]
    7. Device according to one of the preceding claims, the number of families for the bust being between 2 and 6, and the number of families for the legs can be between 2 and 4.
    [8" id="c-fr-0008]
    8. Device according to one of claims 6 and 7, the families for the legs are:
    Shorts
    Short skirt Long skirt
    Trousers.
    5
    [9" id="c-fr-0009]
    9. Device according to one of the preceding claims, a family being provided to describe accessories:
    - No scarf
    - A scarf
    - A hat
    [10" id="c-fr-0010]
    10 - A hat.
    10. Device according to one of the preceding claims, the processing unit being arranged for:
    - determine a confidence rate on the recognition of a
    15 family,
    - when this confidence rate is above a predetermined threshold, for example a 90% threshold, validating the clothing index without passenger intervention,
    - possibly inform the passenger of the family or families
    20 selected to manage their comfort.
    类似技术:
    公开号 | 公开日 | 专利标题
    WO2018229384A1|2018-12-20|Device for identifying a clothing family index
    EP3347237A1|2018-07-18|Thermal management system for a motor vehicle and corresponding thermal management method
    US7961906B2|2011-06-14|Human detection with imaging sensors
    US9723224B2|2017-08-01|Adaptive low-light identification
    US20160350908A1|2016-12-01|Method and system for detecting sea-surface oil
    WO2016038784A1|2016-03-17|Driver state determination apparatus
    WO2018202984A1|2018-11-08|Thermal management system for a motor vehicle passenger compartment
    US20050011959A1|2005-01-20|Tags and automated vision
    FR2917854A1|2008-12-26|CONTROL SYSTEM HAVING A PHYSIOLOGICAL DATA SENSOR FOR AN AIR CONDITIONING INSTALLATION OF A MOTOR VEHICLE.
    JP2007232652A|2007-09-13|Device and method for determining road surface condition
    WO2017016784A1|2017-02-02|Control system for air conditioning system of a motor vehicle
    CN108417043A|2018-08-17|A kind of detection method that high beam persistently opens
    Wong et al.2010|An effective trespasser detection system using thermal camera
    WO2020212366A1|2020-10-22|System for managing the thermal comfort of a person
    FR2958774A1|2011-10-14|Method for detecting object e.g. obstacle, around lorry from stereoscopic camera, involves classifying object from one image, and positioning object in space around vehicle by projection of object in focal plane of stereoscopic camera
    WO2008142238A2|2008-11-27|Method of determining the number of objects present in a given surface region by infrared thermography
    US20210160440A1|2021-05-27|Parallax correction using cameras of different modalities
    FR3104874A1|2021-06-18|Electronic device for identifying and tracking at least one target, associated electronic system, method and computer program
    WO2021170951A1|2021-09-02|System for managing the thermal comfort of a passenger
    CN111179581B|2022-02-01|System and method for sharing bicycle ID matching based on vision
    US20210150234A1|2021-05-20|System and method for cargo management in a motor vehicle
    US11232685B1|2022-01-25|Security system with dual-mode event video and still image recording
    FR3109688A1|2021-10-29|Method of authentication or identification of an individual
    Sherstjuk et al.2019|Evaluation of Fire Intensity Based on Neural Networks in a Forest-Fire Monitoring System
    FR2918237A1|2009-01-02|METHOD FOR NETWORKING CAMERAS OF A NETWORK AND METHOD FOR OPERATING SUCH A NETWORK
    同族专利:
    公开号 | 公开日
    FR3067835B1|2020-05-08|
    US20210232834A1|2021-07-29|
    EP3639202A1|2020-04-22|
    WO2018229384A1|2018-12-20|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US5145112A|1990-10-08|1992-09-08|Kabushiki Kaisha Toyota Chuo Kenkyusho|Air conditioner|
    US5555512A|1993-08-19|1996-09-10|Matsushita Electric Industrial Co., Ltd.|Picture processing apparatus for processing infrared pictures obtained with an infrared ray sensor and applied apparatus utilizing the picture processing apparatus|
    DE102013001878A1|2013-02-02|2014-08-07|Audi Ag|Method for controlling air-conditioning and/or ventilation device of vehicle, invovels determining controlling/regulating parameters of air conditioning and/or ventilation device based on determined climate parameter information|
    DE10257587B3|2002-12-09|2004-04-15|Daimlerchrysler Ag|Climate-control regulation method for automobile with opening roof switches between different regulation modes dependent on open or closed position of opening roof|
    JP2005306095A|2004-04-19|2005-11-04|Denso Corp|Air-conditioning control device for vehicle|FR3095161B1|2019-04-19|2021-03-12|Valeo Systemes Thermiques|A person's thermal comfort management system|
    FR3107669A1|2020-02-28|2021-09-03|Valeo Systemes Thermiques|Passenger thermal comfort management system|
    CN112990362B|2021-04-20|2021-08-20|长沙树根互联技术有限公司|Mine car driving grade recognition model training method and device and terminal equipment|
    法律状态:
    2018-12-21| PLSC| Publication of the preliminary search report|Effective date: 20181221 |
    2019-06-28| PLFP| Fee payment|Year of fee payment: 3 |
    2020-06-30| PLFP| Fee payment|Year of fee payment: 4 |
    2021-06-30| PLFP| Fee payment|Year of fee payment: 5 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1755507|2017-06-16|
    FR1755507A|FR3067835B1|2017-06-16|2017-06-16|APPARATUS FOR IDENTIFYING A CLOTHING FAMILY INDEX|FR1755507A| FR3067835B1|2017-06-16|2017-06-16|APPARATUS FOR IDENTIFYING A CLOTHING FAMILY INDEX|
    US16/622,022| US20210232834A1|2017-06-16|2018-06-01|Device for identifying a clothing family index|
    EP18736987.1A| EP3639202A1|2017-06-16|2018-06-01|Device for identifying a clothing family index|
    PCT/FR2018/051270| WO2018229384A1|2017-06-16|2018-06-01|Device for identifying a clothing family index|
    [返回顶部]