Systems and methods for detecting a seatbelt-wearing state of a passenger of a vehicle

Abstract

According to one aspect, the present description relates to a system (100) for detecting the seatbelt-wearing state of at least one passenger of at least one first vehicle (301), the system comprising detection means configured to acquire, according to a linear and / or circular polarisation, at least one first polarised image of a field of view of the first vehicle, consisting of a plurality of pixels, and determining at least one enriched image with n channels of the field of view of the first vehicle, n ≥ 1, wherein the enriched image consists of a plurality of pixels comprising n values, at least one of which values is a polarisation value determined from pixel values of the at least one first polarised image. The system further determines, from the at least one enriched image, a seatbelt-wearing state of the at least one passenger of the first vehicle.

Description

[0001] Systems and methods for detecting seat belt use by a vehicle passenger

[0002] Technical field of the invention

[0003] The invention relates to systems and methods for detecting the wearing of a seat belt by a passenger in a vehicle, and in particular in a moving vehicle.

[0004] State of the art

[0005] In a context where road safety remains a major concern, it is important to address the persistent problem of not wearing seat belts despite the legal requirement. Data from the French National Interministerial Road Safety Observatory (ONISR) in 2019 revealed that nearly 23% of vehicle fatalities were not wearing seat belts.

[0006] These alarming figures highlight the need to develop a reliable automatic seat belt monitoring system, with a view to encouraging systematic compliance with this fundamental safety measure, with the primary objectives of contributing to the reduction of the number of road accidents and mitigating their severity.

[0007] There is therefore a need to detect and identify with great precision offences relating to the wearing of the seat belt, e.g. the non-wearing, or incorrect wearing of the seat belt by one or more passengers of a vehicle, and in particular of a moving vehicle.

[0008] Seatbelt detection systems exist and are commonly integrated into vehicles, as described in US patent application 2007 / 0195990 [Ref. 1]. However, such in-vehicle systems only provide information about seatbelt use to passengers and can be easily ignored or even deceived. Furthermore, this information is not available to authorities responsible for enforcing seatbelt use and therefore cannot be used for enforcement purposes. A deep learning method using a processed image received from a License Plate Recognition System (LPRS) camera is known from patent application WO 2020 / 076264 [Ref. 2] for the purpose of informing police units about seatbelt violations.

[0009] Patent application WO 2012 / 160251 [Ref. 3] describes a system for detecting passenger seat belt use from a grayscale image of a vehicle. Other known examples of seat belt compliance monitoring systems for detecting violations combine an artificial intelligence method with images from RGB and / or near-infrared (NIR) cameras. These known devices are described, for example, in patent applications US 2016 / 0078306 [Ref. 4] and CN 106709443 [Ref. 5]. Although the devices described above aim to detect whether a vehicle passenger is wearing a seat belt, a higher level of performance is always desirable.In particular, a high level of confidence is sought in the detection of the belt, including in degraded conditions such as, for example and without limitation, a vehicle moving at a sometimes high speed (greater than 100 km / h), detection at night and / or in adverse weather conditions or for vehicles with tinted and / or opaque windows.

[0010] One objective of this description is to propose a new system and method for determining the status of seat belt use by a passenger in a stationary or moving vehicle with a very high level of confidence, including in degraded conditions.

[0011] Summary of the invention

[0012] In this description, the term "include" has the same meaning as "include" or "contain," and is inclusive or open-ended, not excluding other elements not described or depicted. Furthermore, in this description, the terms "approximately" or "substantially" are synonymous with (meaning the same as) a margin of error of 10% or more, for example, 5%, of the respective value.

[0013] According to a first aspect, the present description relates to a system for detecting the wearing of a seat belt by a passenger of at least one first vehicle, said system comprising: detection means configured for the acquisition of at least one first polarized image of a field of view of said at least one first vehicle, the at least one polarized image being made up of a plurality of pixels and being acquired according to a linear and / or circular polarization;a processing unit configured to: determine, from said at least one polarized image, at least one n-channel enriched image of said field of view of said at least one first vehicle, n >1, the enriched image being made up of a plurality of pixels, each pixel comprising n values, at least one of said values ​​being a polarization value determined from pixel values ​​of said at least one first polarized image, determine, from said at least one enriched image, a state of seat belt wearing by said passenger of said at least one first vehicle.

[0014] The applicant demonstrated that such a system could determine the seatbelt wearing status of a vehicle passenger with excellent reliability. This reliability is based on determining, from at least one polarized image of a field of view of said vehicle, the seatbelt wearing status of said passenger.

[0015] For the purposes of this description, an image is defined as a plurality of pixels, each pixel containing one or more values. An n-channel image is an image consisting of a plurality of pixels, each containing n values. The number n is also called the image depth. A two-channel image is called a two-channel image, and a three-channel image is called a three-channel image. In this description, an image whose number of channels is not specified will be understood as a single-channel image. An n-channel image can always be processed as n single-channel images.

[0016] Thus, in the system according to the first aspect, the enriched image determined by the processing unit contains n channels, that is to say, it is made up of a plurality of pixels, each containing n values. The number n is greater than or equal to 1, and corresponds to the depth of the enriched image.

[0017] According to one or more embodiment examples, n > 2 and at least one enriched image is a multilayer image. The applicants have shown that when the enriched image is a multilayer image, the complementary advantages of each channel can be leveraged, thereby improving detection robustness and providing a richer representation of features, thus enhancing performance.

[0018] At least one of the n values ​​of each pixel constituting the enhanced image is a value determined from pixel values ​​of at least one polarized image acquired with linear and / or circular polarization. This value is referred to as the "polarization value" in this description. Thus, the enhanced image contains additional information compared to an image acquired by conventional optical detection methods, such as an intensity camera, which is unpolarized. This additional information is related to at least one polarization value, and its use in determining seatbelt compliance status from the enhanced image helps improve passenger safety.

[0019] Indeed, the applicant has shown that the detection system described in the first aspect can verify that the passenger's seat belt is correctly fastened, even in adverse conditions, both day and night. The applicant has demonstrated that this capability offered by the system described herein stems in particular from the difference between the woven materials from which seat belts are generally made and the clothing worn by passengers.

[0020] Among other advantages, the first-aspect detection system can operate in real time, meaning it can generate a violation notice within a predetermined timeframe from the acquisition of at least one initial polarized image, e.g., less than a few minutes, for example, less than approximately 10 minutes, advantageously less than approximately 5 minutes, or advantageously less than approximately 3 minutes. Furthermore, the first-aspect detection system can guarantee optimal protection of passenger privacy by strictly adhering to data privacy and security standards, for example, by blurring the faces of identifiable individuals in the images. The system can also be compatible with the specific characteristics of all types of vehicles, e.g.light vehicles (LV), light commercial vehicles (LCV), utility vehicles (LCV) and heavy goods vehicles (HGV), and allow control at low and high speed.

[0021] According to one or more embodiment examples, the detection system described in the first aspect is configured to be mounted on fixed and / or mobile equipment. This fixed and / or mobile equipment is external to at least one vehicle, which, compared to known state-of-the-art systems, allows for a detection system that is independent of the vehicle being monitored.

[0022] In one or more embodiments, the seat belt wearing status corresponds to the correct, non-wearing, or incorrect wearing of the seat belt by the passenger in at least one vehicle. In one or more embodiments, the system may simply determine whether the passenger's seat belt is fastened or not. In some embodiments, the system is also capable of detecting incorrect seat belt wear by the passenger. Incorrect wear may, for example, correspond to a state in which the upper part of the seat belt is detected below the passenger's shoulder, instead of above it.

[0023] Of course, the system may not be limited to detecting the seatbelt wearing status of a single passenger in the first vehicle. According to one or more embodiments, the system's processing unit is configured to determine, from at least one enhanced image, the seatbelt wearing status of one or more passengers in the first vehicle, said one or more passengers being selected from the group comprising a driver, a front passenger, a rear passenger, and combinations thereof. According to one or more embodiments, the detection means are configured to acquire at least a first plurality of p polarized images of the field of view of the first vehicle, p > 2, each polarized image being acquired with linear and / or circular polarization.According to one or more embodiment examples, the acquisition of said at least a first plurality of p polarized images is simultaneous or quasi-simultaneous.

[0024] For the purposes of this description, quasi-simultaneous acquisition means that the images are acquired by the detection means within a time interval of less than approximately 5 ms, for example, less than approximately 3 ms. The first plurality of p polarized images can thus include images acquired simultaneously or quasi-simultaneously.

[0025] The detection methods may include one or more optical detectors.

[0026] An optical detector, as defined herein, generally comprises an optical detection surface and a lens including one or more optical elements configured to form an image of a scene on the optical detection surface. Each optical detector includes an optical axis defined by the lens. The field of view of an optical detector is the portion of space that the optical detection surface can capture. The field of view is, for example, an angular field of view, defined by the optical axis and a field angle. The field angle is calculated from the focal length of the lens and the size of the detection surface.

[0027] According to this description, at least one first polarized image acquired by the detection means is acquired with linear and / or circular polarization. In one or more embodiments, the detection means include at least one polarizing filter.

[0028] A polarizing filter configured for acquiring a linearly polarized image may, in a known manner, include a liquid crystal filter. Such a liquid crystal filter includes, by one example, an arrangement of aligned liquid crystals configured to block unpolarized light while allowing polarized light to pass through in a predetermined direction of the filter.

[0029] A polarizing filter configured for acquiring a circularly polarized image may, for example, include a retarder film associated with a linear polarizing filter, which introduces a predetermined phase shift between the polarization components of the light, thus transforming linear polarization into circular polarization. The detection means may include a plurality of polarizing filters as defined above, each polarizing filter being associated with an optical detector included in the detection means or with a part of an optical detector, for example, a sensor, or a pixel of an optical detector sensor.

[0030] According to one or more exemplary embodiments, the detection means configured for acquiring at least a first plurality of p polarized images of the field of view of said first vehicle, p > 2, include a division-of-focal-plane (DoFP) polarimetric camera. A DoFP polarimetric camera is a type of optical sensor capable of capturing polarization information from a scene. In a DoFP polarimetric camera, the image sensor is equipped with a pixel array, where each pixel is subdivided into n subpixels, n > 1, each subpixel being sensitive to a specific polarization of light. This division of the focal plane allows the camera to simultaneously capture several polarization components for each point in the scene. This results in an n-channel polarized image.

[0031] According to one or more exemplary embodiments, the detection means configured for the acquisition of at least a first plurality of p polarized images of the field of view of said first vehicle, p > 2, comprise a plurality of optical detectors, each optical detector being configured for the acquisition of an image polarized according to a different polarization.

[0032] In embodiments where the detection means comprise a plurality of optical detectors, it is possible for the optical detectors to acquire polarized images of the first vehicle with slightly different fields of view. According to one or more embodiments, the system further comprises image registration means for converting the polarized images acquired by the plurality of optical detectors into a plurality of polarized images of the same field of view of said first vehicle.

[0033] According to one or more embodiments, n > 2 and each pixel of said enhanced image comprises a plurality of polarization values. The enhanced image is thus a multi-channel image, each channel corresponding to a polarization value. Each polarization value is determined from pixel values ​​of said at least one polarized image. Each polarization value may correspond directly to a pixel value of a polarized image, and / or result from a calculation performed on pixel values ​​of at least one polarized image. According to one or more embodiments, at least one polarization value is determined from a linear combination of pixel values ​​of two or more of said polarized images.By linear combination of pixel values ​​of polarized images, we understand that the value of each pixel in the enhanced image corresponding to a given point in the field results from this linear combination of the pixel values ​​of said polarized images at the same point in the field. Such a linear combination can, for example, include a sum, difference, and / or multiplication by a constant of pixel values ​​of polarized images.

[0034] According to one or more embodiment examples, said at least one polarization value is selected from the group comprising: a Stokes parameter, a polarization angle, a degree of polarization.

[0035] According to one or more embodiments, the enhanced image is selected from the group consisting of: a two-layer image in which each pixel includes a value for the polarization angle and a value for the degree of polarization; a two-layer image in which each pixel includes a value for the polarization angle or a value for the degree of polarization and an intensity value independent of polarization; and a multi-layer image in which each pixel includes a value for the polarization angle, a value for the degree of polarization, and an intensity value independent of polarization. The Stokes parameters are, as is known, a set of four values, S0, S1, S2, and S3, that describe the polarization state of an electromagnetic wave. The parameters are expressed as a function of the total beam intensity, its degree of polarization, and parameters related to the shape of the polarization ellipse.They allow us to describe unpolarized, partially polarized and totally polarized light.

[0036] One or more Stokes parameters can be calculated in a known manner by linear combination of pixel values ​​from two or more of said polarized images. In exemplary implementations, a Stokes parameter is chosen from S0, S1, or S2, with, for example, S0 = 10 + 190; S1 = 10 - 190; and S2 = 2145 - S0. I0, I45, and 190 are linearly polarized images with polarization angles equal to α, α+45°, and α+90°, respectively, where α ∈ [0, 180°]. Of course, the Stokes parameters can also be determined by linear combination of pixel values ​​from other polarized images.

[0037] The angle of polarization of light, or degree of polarization, is defined for linearly polarized light as the angle of deviation of the axis of vibration from a reference axis. More precisely, when light is linearly polarized, its vibrations occur in a specific plane, along an axis of vibration. The angle of polarization measures the deviation of the axis of vibration from a reference axis, in a plane perpendicular to the plane of vibration.

[0038] The degree of polarization of light is defined as the proportion of polarized light relative to the total light. The degree of polarization thus indicates how polarized light is compared to unpolarized light. Completely unpolarized light will have a zero degree of polarization, while fully polarized light will have a maximum degree of polarization. The degree of polarization is often expressed as a percentage, ranging from 0% (unpolarized) to 100% (fully polarized). The angle of polarization and the degree of polarization can be determined, notably (but not exclusively), using Stokes parameters, particularly S0, S1, or S2.

[0039] According to one or more embodiment examples, n >2 and each pixel of said enriched image includes an intensity value independent of polarization.

[0040] When the enhanced image is a multilayer enhanced image (n > 2), for example, but not limited to, a multilayer enhanced image in which each pixel includes both a polarization-independent intensity value and a polarization value, such as the degree of polarization, the robustness of the detection is increased and the performance improved. Indeed, the first channel, corresponding to a polarization-independent intensity value, can identify structural features (spatial contours, textures), and the second channel, corresponding to a polarization value, such as the degree of polarization, can identify complementary features, such as the physical characteristics of the materials used for the seatbelt. The optical signature of the seatbelt is thus strengthened, and the reliability of the detection is improved.

[0041] In one or more embodiments, the intensity value is determined from the acquisition of a non-polarized image by an optical detector included in the detection means, the optical detector not being associated with a polarizing filter. In one or more embodiments, the intensity value is determined from the pixel values ​​of a plurality of polarized images acquired by the detection means. For example, the intensity value is determined by summing the pixel values ​​of two orthogonal polarized images from among at least one first polarized image acquired by the detection means. In other embodiments, the intensity value is determined by summing the normalized intensities of all the polarized images from among at least one first polarized image.In yet other examples of implementation, the said intensity value is determined by a sum of pixel values ​​of two left-circularly and right-circularly polarized images.

[0042] According to one or more embodiments, the detection means are configured to acquire at least three polarized images of the field of view of said at least one first vehicle, each polarized image being acquired with a different linear polarization. According to one or more embodiments, the polarized images of said at least one first plurality are acquired for polarization angles equal to α, α+45°, and α+90°, respectively, where αe [0, 180°]. According to one or more embodiments, the polarized images of said at least one first plurality are acquired for polarization angles equal to α, α+45°, α+90°, and α+135°, respectively, where αe [0, 180°].

[0043] According to one or more exemplary embodiments, the processing unit is configured to trigger the acquisition of at least one polarized image of the first vehicle's field of view over time, repeatedly, for example, at a given rate. This results in a plurality of acquisitions of at least one polarized image over time, or a "burst" of acquisitions, each acquisition generating at least one enhanced image from which the seatbelt wearing status of a passenger can be determined. The acquisitions in the burst of acquisitions may be spaced at the same predetermined interval, for example, less than approximately 1 minute, advantageously less than approximately 30 seconds, advantageously less than approximately 10 seconds, or advantageously less than approximately 5 seconds.In some implementation examples, the acquisition burst rate can be controlled according to the vehicle's speed. The higher the speed, the shorter the interval between acquisitions; conversely, the lower the speed, the longer the interval between acquisitions. The acquisition burst might consist of, for example, between 3 and 20 acquisitions. Such an acquisition burst makes it possible to further robustly determine the seatbelt wearing status of an individual passenger and / or to determine the seatbelt wearing status of multiple passengers in the vehicle.

[0044] According to one or more implementation examples, the system is configured to control a plurality of vehicles.

[0045] Thus, according to one or more embodiments, the detection means are configured for acquiring at least one second polarized image of a field of view of a second vehicle. According to one or more embodiments, the detection means may include the first optical detector configured for acquiring said at least one polarized image of the first vehicle and at least one second optical detector configured for acquiring at least one second polarized image of a field of view of a second vehicle.

[0046] According to one or more embodiments, when the system is installed on mobile equipment, e.g., an unmarked traffic control vehicle, the detection means can be configured to acquire at least one polarized image of the first vehicle overtaken on the right (by said control vehicle incorporating the detection system) and / or the first vehicle overtaken on the left and / or a crossing vehicle. When the system is part of fixed equipment intended for monitoring multiple lanes, e.g., highway lanes, the system can, in certain embodiments, contain as many optical devices as there are lanes on the highway, for example, 6 optical detectors for a 3x3 lane highway, 4 optical detectors for a 2x2 lane highway.

[0047] According to one or more exemplary embodiments, the system further includes at least one vehicle illumination source. For example, the vehicle illumination source is configured for near-infrared illumination, beyond a predetermined near-infrared value, for example, beyond approximately 850 nm. The illumination source includes, for example, an infrared light source and / or may include a broad-spectrum source with a high-pass near-infrared filter, i.e., allowing only wavelengths above said predetermined near-infrared value to pass through.

[0048] According to one or more embodiments, the system further includes a lux meter configured to measure ambient light intensity. Based on a predetermined light intensity threshold measured by the lux meter, the processing unit can be configured to trigger the activation of the light source.

[0049] In some embodiments, the illumination source is polarized; that is, the light emanating from the illumination source is polarized. For example, the light emanating from the illumination source may be linearly and / or circularly polarized. For instance, the illumination source may produce light in the near-infrared (IR) and include a polarizing filter as defined above with reference to the detection means. Polarized light illumination can further improve seat belt detection. In some embodiments, two enhanced images can be determined almost simultaneously, one with and one without vehicle illumination, and the detection of a seat belt wearing status can take into account an enhanced image resulting from the subtraction between the image with illumination and the image without illumination.According to one or more exemplary embodiments, the system further includes a detection module configured to detect a vehicle entering a control zone and / or a vehicle leaving a control zone, the processing unit being configured to trigger and / or stop the acquisition of said at least one polarized image when the first vehicle enters said control zone and / or leaves the control zone.

[0050] In the case of a detection system mounted on fixed equipment, the control area can be, in embodiment examples, a predetermined control area, for example an area defined by ground detection cables or laser barriers.

[0051] In the case of a detection system mounted on mobile equipment, for example a vehicle, or fixed but configured to be moved, the control zone can be a virtual zone projected onto the ground and defined by optical and / or radar means mounted on the equipment.

[0052] According to one or more embodiment examples, for example when the detection system according to the first aspect is mounted on mobile equipment, for example a car, said detection module may include a vehicle presence sensor, for example using LiDAR light detection and ranging technology or ToF time of flight technology (according to the abbreviations, respectively, of the Anglo-Saxon expressions "Laser Imaging Detection And Ranging", and "Time of Flight").

[0053] According to one or more embodiment examples, the system further includes an identification module for said at least one first vehicle, said identification module being configured for example for the determination of a registration plate for said at least one first vehicle and / or the determination of the type of said at least one first vehicle and / or for the determination of the color of said at least one first vehicle.

[0054] The identification module may, for example, include one or more ANPR (Automatic License Plate Recognition) systems using technology designed to identify vehicle license plates by means of optical character recognition. The identification module may alternatively or additionally include a system for determining the vehicle type and / or the vehicle color. The identification module may utilize at least one initial polarized image of a field of view of the vehicle acquired by the detection means. Alternatively or additionally, the identification module may utilize another image or plurality of images, for example, acquired by a context camera optionally included in the seat belt detection system, according to one or more embodiments of this description.According to one or more embodiments, the processing unit includes an artificial intelligence (AI) module configured to determine, from at least one enhanced image, the seatbelt wearing status of the passenger in the first vehicle. According to one or more embodiments, the AI ​​module includes a neural network. According to a second aspect of this description, one or more embodiments relate to mobile and / or fixed equipment comprising the seatbelt wearing detection system as defined in the first aspect. According to one or more embodiments, the equipment comprises at least one mobile part and one fixed part.

[0055] According to one or more embodiment examples, the mobile equipment which integrates a security wearing detection system according to the first aspect can be an unmarked control vehicle which travels on traffic lanes taking pictures of the vehicles around it.

[0056] According to one or more embodiments, the equipment incorporating a security wear detection system, as described in the first aspect, is fixed. It is configured, for example, to be placed on one or both sides of a road, e.g., on a median strip, a sidewalk, etc. The fixed equipment can also be placed above a road, e.g., attached to a traffic light, a highway overpass, etc. According to one or more embodiments, such fixed equipment can be advantageously configured for easy relocation.

[0057] According to a third aspect, the present description relates to a method for detecting the wearing of a seat belt by at least one passenger of at least one first vehicle by means of one or more examples of the embodiment of a detection system according to the first aspect.

[0058] Thus, one or more examples of embodiment relate to a method for detecting the wearing of the seat belt by at least one passenger of at least one first vehicle, said method comprising: the acquisition of at least one polarized image of a field of view of said at least one first vehicle, the at least one polarized image being made up of a plurality of pixels and being acquired according to a linear and / or circular polarization;the determination, from said at least one polarized image, of at least one n-channel enhanced image of said field of view of said at least one first vehicle, n > 1, the enhanced image being made up of a plurality of pixels, each pixel comprising n values, at least one of said values ​​being a polarization value determined from pixel values ​​of said at least one polarized image, the determination, from said at least one enhanced image, of a state of the wearing of the seat belt by said passenger of said at least one first vehicle.

[0059] According to one or more embodiments, the method further includes a step of illuminating said at least one first vehicle, synchronized with the acquisition step. According to one or more embodiments, the method further includes detecting the entry of said at least one first vehicle into a control zone, with acquisition being triggered when said at least one first vehicle enters the control zone. According to one or more embodiments, the method further includes detecting the exit of said at least one first vehicle from the control zone and stopping acquisition following the exit of said at least one first vehicle.

[0060] Brief description of the figures

[0061] Other features and advantages of the invention will become apparent from the following description, illustrated by the following figures:

[0062] Fig. 1 A represents a diagram of an example of a seat belt wear detection system as described herein;

[0063] Fig. IB represents a diagram of an example of mobile equipment, in this example a car, equipped with an example of a seat belt wear detection system as described herein;

[0064] Fig. 2A represents a diagram of an example of mobile equipment, in this example a control vehicle, seen from above, equipped with an example of a seat belt wear detection system according to this description, and in which several variants for the location of optical detectors of the detection means are illustrated;

[0065] Fig. 2B represents a diagram of an example of mobile equipment, in this example a control vehicle, seen from above, equipped with an example of a seat belt wear detection system according to this description, and in which several variants for the location of optical detectors of the detection means are illustrated;

[0066] Fig. 3 A represents a first diagram illustrating, in top view, the control of the wearing of the seat belt by means of a detection system according to the present description, mounted on a mobile equipment, in this example a control vehicle;

[0067] Fig. 3B represents a second diagram illustrating, from a top view, the control of the wearing of the seat belt by means of a detection system according to this description, mounted on a mobile piece of equipment, in this example a control vehicle; Fig. 3C represents a third diagram illustrating, from a top view, the control of the wearing of the seat belt by means of a detection system according to this description, installed on a fixed piece of equipment;

[0068] Fig. 4 illustrates a step diagram of an example of a seat belt wearing detection process according to this description;

[0069] Fig. 5A represents, by way of illustration, an intensity image of seat belts (polarization independent image), an angle of polarization (AOP) image of the seat belts shown on the intensity image, and a degree of polarization (DOLP) image of the seat belts shown on the intensity image;

[0070] Fig. 5B represents, by way of illustration, an intensity image of a driver of a moving vehicle (polarization-independent image), an angle-of-polarization (AOP) image of the same driver shown in the intensity image, and a degree-of-polarization (DOLP) image of the same driver shown in the intensity image; Fig. 6 illustrates a schematic of a focal-plane-division polarimetric camera adapted for detection means of a seat belt wear detection system as described herein.

[0071] Detailed description

[0072] In the figures, the elements are not shown to scale for better visibility. Fig. 1 A shows a diagram of an example of a seat belt wear detection system according to this description, referenced 100 in Fig. 1 A, and Fig. IB shows a diagram of an example of a mobile device 200, in this example a control vehicle, equipped with a seat belt wear detection system according to this description.

[0073] The system 100 illustrated in Fig. 1A includes detection means 110 configured for acquiring at least one first polarized image of a field of view of a first vehicle controlled by the system 100. The system 100 further includes a processing unit 150 configured to trigger the acquisition of said at least one first polarized image, for example, when said first vehicle enters a predetermined control zone. The processing unit 150 is further configured to determine, from said at least one first polarized image, at least one enhanced image, and to determine, from said at least one enhanced image, the seatbelt wearing status of a passenger in the first vehicle.

[0074] In the embodiment illustrated in Fig. 1A, the detection means 110 comprise one or more optical detectors. According to one or more embodiments, the detection means comprise at least one focal-plane split-polarimetric camera, a schematic of which is shown in Fig. 6. A focal-plane split-polarimetric camera is configured for the simultaneous acquisition of a plurality of images polarized according to different linear polarizations, for example, a plurality of 4 images polarized according to 4 linear polarizations, as described in Fig. 6.

[0075] As illustrated in Fig. 1 A, the system 100 can further include one or more elements selected from an illumination source 120, a lux meter 121, a detection module 130, an interface module 140, a modulator-demodulator 141 in order to provide, for example, internet access to the system, and an electrical power source 142 in order to power the system, e.g., a battery.

[0076] The illumination source 120, intended to illuminate one or more vehicles controlled by the system 100, can be configured for near-infrared illumination beyond a predetermined near-infrared value, for example, beyond approximately 850 nm. The illumination source 120 includes, for example, a high-pass near-infrared filter, i.e., one that only allows wavelengths above this predetermined near-infrared value to pass through. The lux meter 121 is configured to measure the ambient light intensity. Based on a predetermined light intensity threshold measured by the lux meter 121, the processing unit 150 can be configured to trigger the activation of the illumination source 120.

[0077] The 130 detection module can, for example, be configured to detect a vehicle entering and / or exiting a predetermined control zone. The 130 detection module can include one or more vehicle presence sensors. Such sensors can, for example, use LiDAR and / or ToF technology.

[0078] Fig. IB shows a diagram of an example of a mobile device 200, in this example a control vehicle, equipped with a seat belt detection system as described herein. The detection system mounted on the control vehicle 200 shown in Fig. IB may be similar to that shown in Fig. 1A. It includes detection means, including in this example a camera 110 for acquiring polarized images of vehicles being monitored by the vehicle 200, for example, a polarimetric camera. As illustrated in Fig. IB, the camera 110 can be mounted, for example, on a bracket on the roof of the vehicle 200, as can an illumination source 120 and a lux meter 121 as described above.In one embodiment, the system may further include a vehicle detection module including two vehicle presence sensors 130, which can be placed at the front and rear of the vehicle 200 as shown in Fig. IB.

[0079] Generally, the processing modules, processing units, or control units referred to in this description may comprise one or more physical entities and be housed in one or more computers. Where reference is made in this description to calculation or processing steps for the implementation of process steps, it is understood that each calculation or processing step may be implemented by software, hardware, firmware, microcode, or any appropriate combination of these technologies. When software is used, each calculation or processing step may be implemented by computer program instructions or software code.These instructions can be stored or transmitted to a storage medium readable by the processing unit and / or executed by the processing unit to implement these calculation or processing steps. The processing unit 150 can thus, for example, be part of a computer that may also integrate a modem 141. The modem 141 can allow the computer to access online resources and / or resources stored in a cloud or data cloud. An interface module 140 can optionally be integrated into the driver's compartment of the vehicle 200 and provide an interface between an operator and the seat belt detection system. For example, the interface module 140 can, among other things, allow the operator to manually trigger the acquisition of polarized images or context images of vehicles being inspected, or to activate or deactivate the illumination of vehicles by the illumination source 120.

[0080] The system embedded on a mobile device may have its own source of electrical power 142, as illustrated in Fig. 1 A and Fig. IB, or alternatively be powered by the said mobile device, in cases where the latter has a battery.

[0081] Fig. 2A shows a diagram of an example of mobile equipment 200, in this example a control vehicle, viewed from above, equipped with an example of a seat belt detection system as described herein, and illustrating several variations for the placement of optical detectors 110 of the detection means. The optical detectors 110A, 110B, 110C, and 110D are, in this example, cameras configured for acquiring polarized images of the vehicles to be inspected. The detection system may include one or more of the cameras 110A to 110D, whose positions illustrate variations depending, for example, on the driver's seat position, i.e., right-hand or left-hand, which varies by country.As mentioned above, according to one or more embodiments, the system can be configured to detect seat belt use by one or more passengers selected from a group including a driver, a front passenger, a rear passenger, and combinations thereof. Furthermore, the system can be configured to monitor a plurality of vehicles, for example, vehicles being overtaken, vehicles overtaking, and / or vehicles crossing. The number, axis direction, and position of the optical sensors 110 can therefore be predetermined, as required, for example, in light of country-specific regulations. According to one or more embodiments, the axis of the optical sensors is adjustable, for example, during monitoring at high speed. An operator can adjust the axis of at least one optical sensor via, for example, the interface module 140 described above.

[0082] Fig. 2B shows a schematic of an example of a mobile unit 200, in this example a control vehicle, viewed from above, equipped with an example of a seat belt detection system as described herein, and illustrating several variants for the placement of optical detectors 170 of the detection means. Optical detectors 170A and 170B, unlike optical detectors 110A-110D shown in Fig. 2A, are not configured for acquiring polarized images of vehicles to be monitored, but are optionally included in the detection means to acquire contextual images of the potential infraction scene.Such context images, independently of the polarized images acquired by optical detectors 110, or according to one or more variants, in addition to said polarized images, can be used by the detection system to detect and identify monitored vehicles by determining, for example, the license plate, type, and / or color of a monitored vehicle. The detector(s) 170 can, for example, be positioned centrally relative to the width of the monitoring vehicle 200 in order to acquire context images for detecting and identifying monitored vehicles. In the example illustrated in Fig. 2B, two optical detectors 170A and 170B are advantageously positioned at the front and rear of the roof of the monitoring vehicle 200, in order to acquire context images of monitored vehicles passing (or crossing) and overtaken, respectively.

[0083] As illustrated in Figs. 3A and 3B, the seatbelt detection system described herein can, for example, be integrated into an unmarked patrol vehicle 200 that travels on traffic lanes, taking pictures of the vehicles 301 and 302 surrounding it. Alternatively, the detection system can be installed on fixed equipment, as in the specific example of fixed equipment shown in Fig. 3C. In some embodiments (not shown), the detection system can be integrated into both a fixed part and a moving part of the same equipment.

[0084] As indicated above with reference to Fig. 2A, the means of detection may vary, for example depending on the type of equipment integrating the system which is envisaged.

[0085] According to one or more embodiment examples, when mobile equipment is used, for example a control vehicle 200 travelling on one or more traffic lanes as illustrated in Fig. 3 A and Fig. 3B, the detection means of the system may include one or more optical detectors, e.g., among those 110A-110D shown in Fig.

[0086] 2A, positioned to optimize the acquisition of polarized images of fields of view of vehicles 301, 302 in the vicinity of said control vehicle 200.

[0087] Figure 3A shows a first diagram illustrating, from a top view, the control of seat belt use by means of a detection system as described herein, mounted on a mobile unit 200, in this example a control vehicle. In Figure 3A, the control vehicle 200 is traveling on a 2x2 lane highway 30 (31, 32, 33, 34) and is in a configuration that allows it to simultaneously monitor an overtaken vehicle 301 traveling in the slow lane 31 in the same direction as lane 32, and an overtaking vehicle 302 traveling in the fast lane 33 in the opposite direction. Figure 3A also illustrates two control zones 131, 132, delimited by dotted lines. Such zones, in the case of a mobile equipment 200, are relative to the control vehicle 200 in the sense that they vary in space relative to the movement of the mobile equipment.When controlled vehicles 301, 302 enter these areas, the detection system processing unit triggers the acquisition of polarized images of the controlled vehicles' fields of view.

[0088] Fig. 3B shows a second diagram illustrating, from a top view, the control of seat belt use by means of a detection system as described herein, mounted on a mobile device 200, in this example a car. The control vehicle 200 shown in Fig. 3B may be the same as or different from the one shown in Fig. 3A. In Fig. 3B, the control vehicle 200 is traveling on a 3x3 lane highway 32, of which only 4 are shown (31, 32, 33, 34), and is in a configuration that allows it to monitor both an overtaken vehicle 301 and an overtaking vehicle 302, traveling, respectively, in the slow lane 31 and in the fast lane in the same direction as lane 32, which in this example is the middle lane of the 3 lanes (31, 32, 33).

[0089] The optimization of polarized image acquisition of the fields of view of vehicles 301 and 302 near the control vehicle 200 can, for example, be based on the proximity of an optical detector of the detection means to the driver's seat of one or more vehicles being monitored. The shorter the distance between the optical detector and the driver's seat of the vehicle being monitored, the more the reliability of the detection system can be considered optimized for detecting, as a priority, whether a driver is wearing a seatbelt. Of course, the positioning of one or more optical detectors of the detection means can also be optimized to detect, as a priority, whether a passenger other than the driver is wearing a seatbelt.

[0090] In the exemplary case illustrated in Fig. 3 A, two cameras with orientations such as those of optical devices 110B and 110C in Fig. 2A were placed on the roof of the control vehicle 200, in order to optimize the acquisition of images of the driver of overtaken vehicles such as vehicle 301, or crossing vehicles such as vehicle 302.

[0091] In the exemplary case illustrated in Fig. 3B, two cameras with orientations such as the optical devices 110C and 110D of Fig. 2A were placed on the roof of the control vehicle 200, in order to optimize the acquisition of images of the driver of overtaken vehicles such as vehicle 301, or overtaking vehicles such as vehicle 302.

[0092] Of course, the detection methods of the on-board system in the control vehicle 200 may include fewer or more optical devices than those shown in Figs. 2A, 2B, 3A, 3B, and 3C, with identical or different positions. For example, the configurations illustrated in Figs. 3A and 3B, which aim to optimize the detection of seatbelt use by the driver of the vehicles being monitored, are adapted for vehicles with a left-hand drive configuration (as in Germany, France, etc.), and not for vehicles with a right-hand drive configuration (as in India, Japan, the United Kingdom, etc.).

[0093] The system according to the first aspect of this description can also be integrated into fixed equipment, intended for example to be placed on one side, or both sides of a road, e.g., on a median strip, a sidewalk, etc.

[0094] Fig. 3C shows a third diagram illustrating, from a top view, the control of seat belt use by means of a detection system as described herein, installed on a fixed device 400. The example of fixed device 400 shown in Fig. 3C is a device located on a central reservation 34 of a 3x3 lane motorway 30, of which only 3 lanes are shown (31, 32, 33). Positioned in this way, the device 400 can prioritize the control of passengers in a vehicle 302 traveling on the motorway 33. The fixed device can take any form suitable for detecting seat belt use, for example, the form of a radar-type speed control device. The fixed device can be adapted to the specific characteristics of all types of vehicles that can be monitored.For example, in order to monitor heavy goods vehicles (HGVs), at least one optical device included in the system's detection means may be mounted at a minimum height, for example, greater than 2m, greater than 3m, greater than 4m, greater than 5m. The fixed equipment and / or all or part of the system's detection means being integrated therein may, in certain embodiments, be placed above one or more traffic lanes, e.g., attached to a traffic light, a motorway bridge, etc.

[0095] According to one or more embodiments, the fixed equipment can be advantageously configured for easy relocation, for example, to prevent damage or to facilitate maintenance or redeployment in order to prioritize a new area for monitoring. Fig. 4 illustrates a step diagram of an example of a seatbelt wear detection method as described herein.

[0096] The method described in Fig. 4 comprises an acquisition step 402 of at least one polarized image of a field of view of a first vehicle, the at least one polarized image being composed of a plurality of pixels and being acquired with linear and / or circular polarization, for example, when the vehicle enters a predetermined control zone. The method further comprises, in accordance with this description, a determination step 403, from said at least one polarized image, of at least one enhanced image as defined above, and a determination step 406, from said at least one enhanced image, of the seatbelt wearing status of a passenger of the first vehicle. Said processing module 403 may incorporate a software function retrieving part or all of the polarized image(s) made available by the detection means.Such treatment is described in more detail below with reference to Fig. 5A and Fig. 5B.

[0097] As illustrated in Fig. 4, the method may include a detection step 401 of the first vehicle in the control zone. This (optional) step may, for example, be performed by one or more detection modules 130 described above with reference to Figs. 3A and 3B, and configured to detect the entry and / or exit of a vehicle from said control zone. Once produced, at least one enhanced image may be made available to a seat belt status determination module 406. Module 406 may determine, or contribute to determining, from at least one enhanced image, the seat belt status of the passenger in the first vehicle. For example, module 406 may be configured to detect, from at least one enhanced image, a vehicle, a vehicle passenger, and / or a seat belt covering the surface of a vehicle passenger.

[0098] Module 406, for example, integrates an artificial intelligence (AI) module. Module 406 can advantageously include one or more neural networks.

[0099] The neural network(s) can, for example, be convolutional neural networks as described, e.g., in the article by J. Redmon et al. [Ref. 6]. A convolutional neural network is a mathematical function composed of successive layers of sub-functions called neurons. The neural network takes as input a vector consisting of the intensity values ​​of each pixel in an image to be analyzed. The network can successively apply filtering (or "activation") operations to all the pixels provided to it. The activation of each neuron determines a "weight" that has been previously determined through learning; a weight is a synaptic coefficient that establishes contact between two neurons above a certain threshold.The weight can then determine a value of the image associated with a label, for example: "seat belt", "vehicle", "truck", "passenger", "van", "vehicle brand", "passenger shoulder", "driver", "child", etc.

[0100] In one or more implementation examples, the weights of the neurons in the network are determined through supervised learning of the neural network. This learning involves feeding the network a large quantity of labeled images, for example, thousands or millions, allowing it to adjust these weights. In one or more implementation examples, the AI ​​module uses a deep learning technique, an algorithm employing image processing techniques, solid-state scanning (SSD), multiple convolutional neural networks, and combinations thereof.

[0101] Thus, in some implementation examples, the network can successively perform calculations on all the provided values. Each neuron calculates the features using "weights" previously determined through training, then feeds all the information associated with the features into the feature fusion layer and classifies them. The weights can then determine the image values ​​associated with the labels, for example, "person," "person wearing a seatbelt," "person not wearing a seatbelt," and so on. In one implementation example, as described earlier, the network's weights are determined through "supervised" network training. This training involves feeding thousands of labeled images into the network to allow it to adjust the weights.In this way, the step of determining whether a seat belt is worn or not in each enhanced image of each enhanced image series determines one or more people and the coordinates of those people in the images and whether they are wearing a seat belt or not.

[0102] When the enriched image is a multilayer enriched image (n > 2), for example and without limitation, a multilayer enriched image in which each pixel includes an intensity value independent of polarization and a polarization value, for example a degree of polarization value, the artificial intelligence module will be able to operate in the following way.

[0103] In some implementation examples, the artificial intelligence module can be configured to process as input the combined values ​​of the multilayer enriched image ("early fusion"), for example, the polarization-independent intensity value and the polarization value. In these examples, a multichannel input tensor can be determined from the polarization-independent intensity value and the polarization value. The values ​​can be normalized beforehand. The input tensor can be fed into a convolutional neural network (CNN) and into an object detection infrastructure, for example, a CSPDarknet53 neural network using DarkNet-53. The neural network can automatically learn the representation of multichannel features for seatbelt detection.

[0104] In implementation examples, the artificial intelligence module can be configured for processing based on features determined from the multilayer enriched image ("late fusion").

[0105] In these examples, a convolutional neural network (CNN) can be used, for example, a dual-branch CNN, to extract spatial or texture features from the intensity value image and physical material characteristics from the polarization value image. These features are combined by an attention mechanism (for example, a convolutional block attention module, or CB AM) for use by a detection module. This enables object detection, including seatbelt detection, for example, an anchor-free detector that combines a feature pyramid network (FPN) and a flexible keypoint detector.

[0106] In all cases, such multilayer images make it possible to take full advantage of the complementary benefits of each channel, thus improving the robustness of detection and providing a richer representation of features, which improves performance.

[0107] The detection method according to this description may optionally include an acquisition step 404 of at least one background image, as illustrated in Fig. 4. The at least one background image may, for example, be acquired by one or more optical detectors, such as optical detectors 170A and 170B illustrated in Fig. 2B, included in the detection means but independent of optical detector(s) configured for the acquisition 402 of the at least one polarized image, e.g., detectors 110A to 110D illustrated in Fig. 2A.

[0108] At least one background image can be made available to a license plate recognition module 405, for example, an ANPR (Automatic Number Plate Recognition) module, as illustrated in Fig. 4. In the illustrated method example, the ANPR module 405 also uses at least one enhanced image. Alternatively or additionally, the ANPR module 405 can also use one or more polarized images from among the at least one polarized image acquired by the detection means.

[0109] More generally, the process may include a 405 identification module comprising one or more elements selected from an ANPR module, a vehicle type determination system, a vehicle color determination system and / or other vehicle characteristics, and combinations thereof.

[0110] In the method example shown in Fig. 4, the ANPR module 405 detects and extracts a license plate from the first vehicle, e.g., its license plate number. In the method example shown in Fig. 4, the information extracted from the AI ​​modules 406 and ANPR module 405 is made available to a violation message generation module 407. This violation message generator 407 can be configured to generate a violation message containing evidence and / or images of the violation, as well as a means of identifying the offending vehicle, e.g., its license plate number. In some embodiments, the generation module 407 receives vehicle position identification data, e.g., GPS data, provided by the vehicle detection module 401.

[0111] The resulting violation message can be sent, for example, to a 408 monitoring system, e.g., a monitoring system for control equipment. The monitoring system can be independent, or it can be part of the detection system or the equipment incorporating the detection system.

[0112] Fig. 5 A represents, by way of illustration, an intensity image 51 of seat belts (polarization independent image), an angle of polarization (AOP) image 52 of the seat belts shown in intensity image 51, and a degree of polarization (DOLP) image 53 of the seat belts shown in intensity image 51.

[0113] Fig. 5B illustrates an intensity image 54 of a driver in a moving vehicle (polarization-independent image), an angle-polarization (AOP) image 55 of the same driver shown in the intensity image 54, and a degree-polarization (DOLP) image 56 of the same driver shown in the intensity image. In some embodiments, the detection system detects incorrect seat belt use by the passenger. Incorrect use might, for example, correspond to a state in which the upper part of the seat belt is detected below the passenger's shoulder, instead of above it.

[0114] The detailed description below gives two illustrative examples of the determination of images in AOP or images in DOLP such as those 52, 55 (AOP) and 53, 56 (DOLP) represented in Fig. 5 A and Fig. 5B.

[0115] According to one or more embodiments, the detection means are configured to acquire three polarized images of the field of view of the first vehicle, each polarized image being acquired with linear polarization. The polarized images (I0, I45, I90) of at least one first plurality can, for example, be acquired for polarization angles equal to α, α+45°, and α+90°, respectively, where αe [0, 180[. According to one or more embodiments, the processing unit is then configured to determine, from the polarized images (I0, I45, I90), at least three images of the Stokes parameters (S0, S1, S2), where S0 = I0 + I90; S1 = I0 - I90; and S2 = 2I45 - S0. The Stokes parameter S0 in these embodiments corresponds to the intensity value S0 defined above.At least one polarization value among the n values ​​of each pixel constituting the enhanced image can then be determined from the Stokes parameter images. According to one or more embodiments, this at least one polarization value is a polarization angle and / or a degree of polarization. For example, the degree of polarization.

[0116] (DOLP) can be determined according to the following formula (1):

[0117] [Math 1] in which DOLP G [ 0, 1],

[0118] As for the polarization angle (AOP), it can, for example, be determined according to the following formula (2):

[0119] [Math 2]

[0120] S2 AOP = arctan —

[0121] SI in which AOP G [ - 90°, 90°].

[0122] According to one or more exemplary embodiments, the detection means are configured to acquire four polarized images of the field of view of the first vehicle, each polarized image being acquired with linear polarization. The polarized images (I0, I45, I190, I135) of at least one first plurality can, for example, be acquired for polarization angles equal to α, α+45°, α+90°, and α+135°, respectively, where αe ∈ [0, 180°[. According to one or more exemplary embodiments, the processing unit is then configured to determine, from the polarized images (I0, I45, I190, I135), at least three Stokes parameters (S0, S1, S2), where S0 = I0 + I90; S1 = I0 - I90; and S2 = I45 - I135. The Stokes parameter S0 in these exemplary embodiments corresponds to the intensity value S0 defined above.At least one polarization value among the n values ​​of each pixel constituting the enhanced image can then be determined from the Stokes parameters. In these embodiments, when said at least one polarization value includes or consists of a polarization angle (AOP) and / or a degree of polarization (DOLP), these values ​​can also be determined according to the mathematical formulas (1) and (2) given above. Fig. 6 illustrates a schematic of a focal plane-splitting polarimetric camera adapted for detection means in a seat belt detection system as described herein. The at least one focal plane-splitting polarimetric camera can be schematically represented as shown in Fig. 6. Only a portion of the camera is shown in Fig. 6.In general, in a focal plane division polarimetric camera (DoFP), each pixel 60 of a camera detection area arranged in a focal plane contains the same number and arrangement of polarizers with different orientations, for example, four polarizers with different polarization angles, as illustrated in Fig. 6: 0°, 45°, 90°, and 135°. This allows the polarization state of the incident light to be determined. The camera sensor shown in Fig. 6 comprises four identical pixels 60, and therefore allows the simultaneous acquisition of four polarized images 601, 602, 603, and 604 of the field of view of the first vehicle, each with a different linear polarization. These robust and easy-to-use cameras allow for instantaneous polarimetric image acquisition. Furthermore, since the polarizing filters are integrated directly onto the sensor, their properties are very stable, simplifying calibration.In the example illustrated in Fig. 6, the polarizing filters associated with each pixel of the camera sensor are linear polarizers. Although described through a number of exemplary embodiments, the system and method for detecting seat belt use according to this description include various variants, modifications, and improvements that will be obvious to those skilled in the art, it being understood that these various variants, modifications, and improvements form part of the scope of the invention as defined by the following claims.

[0123] References

[0124] Ref. 1: US 2007 / 0195990

[0125] Ref. 2: WO 2020 / 076264

[0126] Ref. 3: WO 2012 / 160251

[0127] Ref. 4: US 2016 / 0078306

[0128] Ref. 5: CN 106709443

[0129] Réf. 6 : J. Redmon et al., “You Only Look Once: Unified, Real-Time Object Detection”, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) (27-30 June 2016), Section 2 “Unified Detection” DOI: 10.1109 / CVPR.2016.91

[0130] 1

Claims

DEMANDS 1. System (100) for detecting the wearing of seat belts by at least one passenger of at least one first vehicle (301), the system comprising: detection means configured for the acquisition of at least one first polarized image of a field of view of said at least one first vehicle (301), the at least one first polarized image being made up of a plurality of pixels and being acquired according to linear and / or circular polarization; a processing unit (150) configured to: trigger the acquisition of said at least one first polarized image;determine, from said at least one first polarized image, at least one n-channel enriched image of said field of view of said at least one first vehicle (301), n ​​>1, the enriched image being made up of a plurality of pixels, each pixel comprising n values, at least one of said values ​​being a polarization value determined from pixel values ​​of said at least one first polarized image, determine, from said at least one enriched image, a state of the wearing of the seat belt by said passenger of said at least one first vehicle (301).; 2. System according to claim 1, wherein: the detection means (110) are configured for the simultaneous or quasi-simultaneous acquisition of at least a first plurality of p polarized images (601 - 604) of the field of view of said at least a first vehicle, p > 2, each polarized image being acquired according to a linear and / or circular polarization.

3. System according to claim 2, wherein at least one polarization value is determined from a linear combination of pixel values ​​of two or more of said polarized images (601 - 604).

4. System according to claim 3, wherein said at least one polarization value is selected from the group comprising: a polarization angle, a degree of polarization.

5. A system according to claim 4, wherein said enhanced image is selected from the group consisting of: a bilayer image in which each pixel comprises a value of the polarization angle and a value of the degree of polarization; a bilayer image in which each pixel comprises a value of the polarization angle or a value of the degree of polarization and an intensity value independent of polarization; and an image multilayer in which each pixel includes a value of the polarization angle, a value of the degree of polarization, and an intensity value independent of the polarization.

6. System according to any one of claims 2 to 5, wherein the detection means are configured for the simultaneous acquisition of 4 polarized images of the field of view of said at least one first vehicle (301), according to 4 different linear polarizations, for example, the polarized images (Io, I45, 190, 1135) of said at least one first plurality are acquired for polarization angles equal to respectively a, a+45°, a+90° and a+135° in which ae [0, 180[.

7. System according to any one of claims 2 to 6, wherein the detection means comprise a focal plane-splitting polarimetric camera (1100), configured for the acquisition of said at least a first plurality of p polarized images.

8. System according to any one of the preceding claims, wherein n > 2.

9. System according to claim 8, wherein each pixel of said enhanced image comprises a plurality of polarization values ​​each determined from pixel values ​​of said at least a first polarized image.

10. System according to any one of claims 8 or 9, wherein each pixel of said enhanced image comprises an intensity value (So) independent of polarization.

11. System according to any one of the preceding claims, wherein the detection means are configured for the acquisition of at least one second polarized image of a field of view of a second vehicle (302).

12. System according to any one of the preceding claims, wherein the system comprises a detection module (130) configured to detect an entry and / or exit of said at least one first vehicle (301) from a control zone (131), the processing unit being configured to trigger and / or stop the acquisition of said at least one polarized image when said at least one first vehicle (301) enters said control zone (131) and / or exits the control zone.

13. System according to any one of the preceding claims, wherein the processing unit (150) is configured to trigger the acquisition of said at least one polarized image of the field of view of said at least one first vehicle over time, repeatedly, for example with a given rate.

14. System according to any one of the preceding claims, further comprising at least one polarized light illumination source (120) of said at least one first vehicle (301).

15. System according to any one of the preceding claims, further comprising an identification module (405) for said at least one first vehicle, said identification module being configured for example for the determination of a registration plate for said at least one first vehicle and / or the determination of the type of said at least one first vehicle and / or for the determination of the color of said at least one first vehicle.

16. System according to any one of the preceding claims, wherein the processing unit (150) comprises an artificial intelligence (AI) module configured to determine, from at least one enhanced image, the state of seat belt wearing by said passenger of said at least one first vehicle (301).

17. Mobile (200) and / or fixed (400) equipment comprising the seat belt wearing detection system (100) as defined in any one of the preceding claims.

18. A method for detecting seat belt use by at least one passenger of at least one first vehicle (301), said method comprising: acquiring (402) at least one polarized image of a field of view of said at least one first vehicle (301), the at least one polarized image being composed of a plurality of pixels and being acquired with linear and / or circular polarization; determining (403), from said at least one polarized image, at least one n-channel enhanced image of said field of view of said at least one first vehicle (301), n ​​> 1, the enhanced image being composed of a plurality of pixels, each pixel comprising n values, at least one of said values ​​being a polarization value determined from pixel values ​​of said at least one polarized image; determining (406), from said at least one enhanced image, the status of seat belt use by said passenger of said at least one first vehicle (301).

19. Method according to claim 18, further comprising a polarized light illumination step of said at least one first vehicle, synchronized with the acquisition step (402).

20. Method according to claim 18 or 19, further comprising the detection of an entry and / or exit of said at least one first vehicle (301) from a control zone (131), and the triggering and / or stopping of the acquisition of said at least one polarized image when said at least one first vehicle (301) enters said control zone (131) and / or exits the control zone.

Images