Method for processing a plurality of images acquired by a camera on board a specific vehicle

Abstract

The invention relates to a method (100) for processing a plurality of images acquired by a camera on board a specific vehicle, each of the images from the plurality of images being associated with respective acquisition times; the method can be used, in particular, to determine (160) a priority vehicle detection index indicating the probability of a vehicle detected in images from the plurality of images being a priority vehicle.

Description

Description Title: Method for processing a plurality of images acquired by a camera mounted on a specific vehicle Technical field

[0001] This disclosure relates to the field of vehicle driving assistance functions. Prior art

[0002] In the context of vehicle driving assistance functions, and in particular in the context of autonomous vehicle driving functions, it is advantageous to be able to detect priority vehicles in circulation, in particular to facilitate their passage.

[0003] A priority vehicle may, for example, be a police vehicle, a medical vehicle such as an ambulance or a light medical vehicle, a fire engine, a construction vehicle, etc. These vehicles are equipped with flashing emergency lights and possibly sirens when they are in operation to warn other drivers of their presence.

[0004] US2019101928 describes a computing device of a first vehicle capable of receiving a first image and a second image of a second vehicle having flashing light signals. The computing device may determine, in the first image and the second image, an image region that bounds the second vehicle such that the image region substantially encompasses the second vehicle. The computing device determines a polar grid that divides the image region in the first image and the second image into polar bins, and identifies portions of image data having a color change and a brightness change between the first image and the second image. The computing device may determine a type of flashing light signals and a type of the second vehicle; and accordingly provides instructions for controlling the first vehicle.

[0005] The solution described in this document requires partitioning the bounding box representing the target vehicle with a polar system, which is limiting. Furthermore, considering the entire bounding box also has the disadvantage of risking returning a large number of false positives, particularly related to complex cases such as: - reflections from the sun which can cause changes in brightness or color from one cycle to another; - some emergency lights are amber in some countries and have flashing frequencies similar to lane change lights.

[0006] Document W02022079113A1 describes a method for processing a video stream of images captured by a color camera to detect a priority vehicle by performing a step of acquiring a sequence of images, then, for each image in the sequence of images, a step of color segmentation by thresholding allowing the detection of colored light zones, a step of tracking each segmented light zone, a step of classifying the colors of each segmented light zone, a step of frequency analysis of each segmented light zone allowing the determination of a flashing character of the zone, a step of calculating an overall reliability index for each image in the sequence of images, allowing the identification of a light zone such as a flashing light.

[0007] Although effective, this solution has limitations for detecting non-stationary flashing lights such as those on an emergency vehicle. When a flashing light is off, it is not measured by the segmentation, and inefficient tracking does not allow for the frequency analysis necessary for detecting the rotating light.

[0008] The present disclosure presents a solution to assist in the detection of this type of vehicles during their intervention free from the aforementioned drawbacks. Summary

[0009] In this regard, a method is proposed for processing a plurality of images acquired by a camera on board a specific vehicle, each image of the plurality of images being associated with a respective acquisition time, the method comprising the following operations: - detecting at least one vehicle traveling in the same direction as the specific vehicle from the plurality of acquired images; - determining, for a detected vehicle, a plurality of areas of interest of the vehicle shared by images of the plurality of images and being likely to include at least one flashing emergency light of a priority vehicle; - detecting the lights present on the images of the plurality of images; - construct a history indicating, for each image of the plurality of images, a presence or absence of a light in each of the areas of interest from the lights detected; - determining, for at least one area of ​​interest, an area confidence index from the information associated with the area of ​​interest concerned listed in the history; and - determining a priority vehicle detection index indicating a probability that the detected vehicle is a priority vehicle from an area confidence index of at least one area of ​​interest of the detected vehicle.

[0010] Optionally, determining an area confidence index of an area of ​​interest includes, when light of the same color has been detected on several images of the plurality of images as comprised at least in part in the area of ​​interest, determining at least one characteristic from: - a detected light flashing frequency; - a variance in flickering of the detected light; - an average duration of the phases during which the detected light is on; or - an average duration of the phases during which the detected light is off.

[0011] According to this option, the confidence index associated with an area of ​​interest of the detected vehicle is determined from at least one characteristic.

[0012] Optionally, an area confidence index is associated with a specific color, such that a specific area of ​​interest is associated with a plurality of confidence indices when a plurality of lights of different colors have been respectively detected on several of the plurality of images as present in the area of ​​interest.

[0013] Optionally, the method further comprises the following operations: - determine, for a detected vehicle, at least one region of interest comprising several areas of interest; - determining, for each region of interest, a region confidence index, the region confidence index of a region of interest being determined from at least one area confidence index of an area of ​​interest belonging to the region of interest.

[0014] According to this option, a priority vehicle detection index is determined from a region confidence index of at least one region of interest.

[0015] Optionally, the method further comprises the following operation: - classify an area of ​​interest as active or inactive based on a value of a confidence index associated with it.

[0016] In this option, a region confidence index of a region of interest is also determined from a weight calculated based on at least one of: a) a positioning of the areas of interest classified as active in a region of interest; or b) a color associated with the areas of interest classified as active in a region of interest.

[0017] Optionally, a specific detected vehicle is a car or a truck; and the at least one specific region of interest determined for the specific detected vehicle comprises at least one of: a) a first region of interest corresponding to an upper portion of the specific vehicle positioned at roof level in the image; or b) a second region of interest corresponding to a central strip of the specific vehicle and extending between the two front lights of the specific vehicle.

[0018] Optionally, the second region of interest is devoid of the vehicle's turn signals.

[0019] Optionally, a specific detected vehicle is a motorcycle and the plurality of areas of interest of the motorcycle comprises at least one area of ​​interest from: a) a first area of ​​interest corresponding to a first side band extending from a first end of the windshield of the motorcycle and comprising a first rearview mirror of the motorcycle; b) a second area of ​​interest extending between two ends of the windshield of the motorcycle located on the same horizontal plane; or c) a third area of ​​interest corresponding to a second side band extending from a second end of the windshield of the motorcycle and comprising a second rearview mirror of the motorcycle.

[0020] Optionally, when the priority vehicle detection index is greater than a predetermined detection threshold for a detected vehicle, the method further comprises the following operation: - identify the detected vehicle as corresponding to a priority vehicle.

[0021] The application also relates to a data processing device configured to implement any of the methods described in the present disclosure.

[0022] The application further relates to a computer program product comprising instructions for implementing any of the methods presented by the present disclosure when this program is executed by a processor.

[0023] Finally, the application relates to a non-transitory recording medium readable by a computer on which is recorded a program for implementing any of the methods presented by the present disclosure when this program is executed by a processor.

[0024] The method presented by the present disclosure thus makes it possible to determine an index indicating a probability that a vehicle detected by a camera on a vehicle is a priority vehicle. It also makes it possible, in certain examples, to determine whether or not the detected vehicle is a priority vehicle. The method is particularly advantageous for driving assistance functions since these functions can use this information to act accordingly, and in particular to facilitate the passage of detected priority vehicles where appropriate. Furthermore, to the extent that the method relies on pre-existing vehicle detection and light detection functions, its implementation does not require significant modification of the consequent software architecture present in vehicles using driving assistance functions and / or autonomous driving functions. Brief description of the drawings

[0025] Other features, details and advantages will become apparent upon reading the detailed description below, and upon analyzing the attached drawings, in which:

[0026] [Fig. 1] Figure 1 schematically represents an example of a data processing device that can be configured to implement a method for processing a plurality of images acquired by a camera on board a vehicle.

[0027] [Fig. 2] Figure 2 represents an example of a flowchart of a method for processing a plurality of images acquired by a camera on board a vehicle.

[0028] [Fig. 3a] Figure 3a schematically represents a view of the front of an ambulance.

[0029] [Fig. 3b] Figure 3b schematically represents a view of the front of a motorcycle.

[0030] [Fig. 4] Figure 4 shows examples of patterns of areas of interest to which a weight can be associated. Description of the embodiments

[0031] A data processing device 10 is now described with reference to FIG. 1, configured to implement the processing operations described below on the images acquired by a camera. The processing device 10 may, for example, comprise a computer 11 enabling the implementation of all or part of the processing operations described below and a memory 12 storing the code instructions executed by the computer.

[0032] The computer 11 may for example be of the processor, microprocessor, microcontroller, FPGA, etc. type. The memory 12 may for example comprise a ROM (Read-Only Memory), a RAM (Random Access Memory), an EEPROM (Electrically Erasable Programmable Read-Only Memory) or any other type of suitable storage means. The memory may for example comprise optical, electronic or magnetic storage means. It may for example be adapted to store the images acquired by the camera.

[0033] The camera capturing the images is mounted on a vehicle (not shown). It may be a rear camera of the vehicle, for example the vehicle's reversing camera. The data processing device 10 may also be mounted on the vehicle with the camera or be arranged remotely from the vehicle. In this second case, the vehicle comprises wireless communication means adapted to communicate the images acquired by the camera mounted on the vehicle.

[0034] With reference to FIG. 2, an example method 100 for processing a plurality of images acquired by a camera mounted on a specific vehicle is presented.

[0035] The image processing method according to the present disclosure makes it possible to determine a priority vehicle detection index indicating a probability that a vehicle detected in the images acquired by the camera is a priority vehicle.

[0036] As explained above, a priority vehicle may, for example, correspond to a police vehicle, a medical vehicle such as an ambulance or a light medical vehicle, a fire engine, a construction machine, etc. A priority vehicle according to the present disclosure is in particular provided with flashing emergency lights.

[0037] As illustrated by FIG. 2, the method 100 comprises an operation 110 of detecting at least one vehicle traveling in the same direction as the specific vehicle from the plurality of images acquired by the camera on board the specific vehicle.

[0038] Algorithms for detecting moving vehicles are known to those skilled in the art. An example of one of these algorithms is described in Sultana, Fahmida & Naimur, S & Amin, Md. (2022). Traffic Participants Detection and Classification Using YOLO Neural Network. 09-18. 10.5281 / zenodo.6844336.

[0039] This type of algorithm generally makes it possible to delimit the vehicles detected on the images by BB limitation frames (generally known as "bounding boxes"), for example rectangular in shape, well known to those skilled in the art. In the case where the camera having acquired the images is a camera located at the rear of the vehicle, the detected vehicles traveling in the same direction as the specific vehicle are visible from their front face. With reference to figures 3a and 3b, an example of a BB limitation frame surrounding respectively the front face of an ambulance and the front face of a motorcycle is shown which could for example be detected by the method 100 during the operation 110.

[0040] As illustrated by FIG. 2, the method 100 comprises an operation 120 of determining a plurality of areas of interest Z r on a vehicle detected following operation 110. Areas of interest Z rare areas of the detected vehicle present on several images of the plurality of acquired images and advantageously present on each of these images. An area of ​​interest Z r designates an area likely to include at least one flashing emergency light of a priority vehicle and preferably exactly one flashing emergency light. The operation 120 of determining a plurality of areas of interest Z r on a vehicle detected following operation 110 can advantageously be carried out for each vehicle detected.

[0041] The position of the plurality of areas of interest Z rdepends on the type (truck, car, motorcycle, etc.) of the vehicle detected in step 110. To this extent, the operation 110 of detecting at least one vehicle may comprise a classification of the at least one vehicle detected by type. This classification may in particular be carried out from a height of the vehicle, a width of the vehicle, a number of wheels of the vehicle, classification algorithms using neural networks or machine learning, etc.

[0042] In first examples in which the detected vehicle corresponds to a car or a truck, the plurality of areas of interest Z r may include at least one of: - a first plurality of zones of interest ZRI positioned at the roof of the detected vehicle, advantageously at the front of the vehicle; and / or - a second plurality of zones of interest ZR2 positioned in a central band of the detected vehicle extending between the two front lights of this vehicle.

[0043] These first examples are notably represented in Figure 3a schematically illustrating a front view of an ambulance truck in which the first plurality of zones of interest ZRI is identified by the references ranging from Z1 RI to Z5RI and the second plurality of zones of interest ZR2 is identified by the references Z1 R2 and Z2R2. Utility vehicles and light medical vehicles are considered to correspond to the type of vehicle “car” in the present application.

[0044] In second examples in which the detected vehicle corresponds to a motorcycle, a plurality of areas of interest Z r may include at least one of: - a first zone of interest ZR micorresponding to a first side strip extending from a first end of the windshield of the motorcycle and comprising a first rearview mirror of the motorcycle, - a second area of ​​interest Zp m 2 extending between two ends of the motorcycle windshield located on the same horizontal plane, - a third area of ​​interest ZR m 3 corresponding to a second side strip extending from a second end of the windshield of the motorcycle and comprising a second rearview mirror of the motorcycle.

[0045] These second examples are notably represented in figure 3b, which schematically illustrates a motorcycle seen from the front.

[0046] As illustrated in Figure 2, the method 100 comprises an operation 130 of detecting the lights present on the images of the plurality of images, and advantageously lights having a color corresponding to a flashing emergency light. Known algorithms make it possible to detect lights on the images. In particular, a light can for example be identified on an image from the colorimetry of the pixels of the image. For example, it can be predefined, in a colorimetric space, intervals of pixel values ​​associated with a red light, a blue light, an orange light, or any light that can correspond to a light having the color of a flashing emergency light. In this way, pixels having this colorimetry in an image can be identified as belonging to a light on the image.A light on an image associated with a specific color can for example correspond to all the neighboring pixels of an image presenting a value included in the range of values ​​associated with this specific color.

[0047] As illustrated in FIG. 2, the method 100 comprises an operation 140 of constructing a history indicating, for each image of the plurality of images, a presence or an absence of a light in each of the zones of interest Z r determined for the detected vehicle. The operation 140 may for example comprise an identification 141, in each image of the plurality of images, of the lights included at least in part in a zone of interest Z rof the detected vehicle then an association 142 of this light with this area of ​​interest in the history. In particular, the color of the light associated with the area of ​​interest and the image can be indicated in the history. This operation involves determining which lights are included at least in part in an area of ​​interest Z r of the detected vehicle likely to include at least one flashing emergency light of the vehicle in order to determine whether an area of ​​interest may be similar to a flashing emergency light, thus making it possible to determine whether the detected vehicle may be a priority vehicle. Thus, the history obtained indicates for each area of ​​interest, the presence or absence of light for each image, and possibly the color of the light present in the area on the images.

[0048] In examples where the pixels composing a light are included in several areas of interest Z rof the vehicle on an image, the light can be associated with the area of ​​interest Z r comprising a central pixel of light or can be associated with the area of ​​interest Z r comprising the largest number of pixels composing the light among the areas of interest Z r comprising pixels of light.

[0049] As illustrated in Figure 2, the method then comprises a determination operation 150, for at least one area of ​​interest Z r , of an area confidence index from the information associated with the area of ​​interest concerned listed in the history. An area confidence index can advantageously be determined for each area of ​​interest Z r determined for the detected vehicle.

[0050] In examples, determining 150 an area confidence index of an area of ​​interest Z rincludes, when light of the same color has been detected on several images of the plurality of images as included at least in part in the area of ​​interest Z r , a determination 151 of at least one characteristic among: - a detected light flashing frequency; - a variance in flickering of the detected light; - an average duration of the phases during which the detected light is on; or - an average duration of the phases during which the detected light is off.

[0051] In these examples, an area confidence index associated with an area of ​​interest Z r of the detected vehicle is determined from at least one characteristic obtained during operation 151.

[0052] These characteristics make it possible to determine the behavior of the lights associated with an area of ​​interest Z r, and in particular their possibly flashing nature. Indeed, by definition, a flashing emergency light is not always on, so that a light will not be detected on the images captured by the on-board camera for which the flashing light is off. Thus, by analyzing in particular the frequency of switching on and off of a zone of interest Z r , it is possible to determine a confidence index reflecting a probability that the area of ​​interest includes a flashing emergency light.

[0053] In examples, when no light is detected as being at least partly within an area of ​​interest Z r on the plurality of images, the area confidence index associated with the area of ​​interest Z r can be determined to be equal to 0.

[0054] In examples, an area of ​​interest Z ris associated with a plurality of confidence indices when a plurality of lights of different colors have been detected respectively on several images of the plurality of images as present in the area of ​​interest. In these examples, each confidence index of the plurality of confidence indices associated with an area of ​​interest Z r specific is associated with a respective color among the plurality of different colors of the lights detected in the area of ​​interest Z r specific.

[0055] As illustrated in FIG. 2, the method comprises an operation 160 of determining a priority vehicle detection index indicating a probability that the detected vehicle is a priority vehicle from an area confidence index of at least one area of ​​interest Z rof the detected vehicle. Advantageously, the priority vehicle detection index can be determined from each of the zone confidence indices determined at the end of operation 150.

[0056] Furthermore, operations 120 to 160 can be carried out for each vehicle detected following operation 110.

[0057] The method according to the present disclosure thus cleverly combines pre-existing software functions for detecting lights on an image and for detecting vehicles by image analysis in order to determine an index indicating a probability that a detected vehicle is a priority vehicle.

[0058] The method presented therefore responds to a practical driving assistance situation, particularly advantageous in the context of autonomous driving, which requires detecting priority intervention vehicles circulating near a vehicle in question, for example to facilitate their passage, and thus guarantee road safety.

[0059] In examples, the method may further comprise an operation 161 of identifying the detected vehicle as corresponding to a priority vehicle when the priority vehicle detection index is greater than a predetermined detection threshold.

[0060] In examples, at least one area of ​​interest Z r , and advantageously each of the areas of interest Z r, of a detected vehicle is devoid of at least one of: a headlight of the detected vehicle and a turn signal of the detected vehicle. This makes it possible to increase the reliability of the priority vehicle detection index determined at the end of the method 100 for a detected vehicle. Indeed, the turn signals of the detected vehicle could be considered as a flashing emergency light and thus reduce the robustness of the method.

[0061] In examples, at least one region of interest R may be determined for a detected vehicle. A region of interest R may comprise or be composed of several areas of interest Z r , as illustrated in Figure 3a in which a first region of interest R1v comprises the first plurality of areas of interest ZRI while a second region of interest R2v comprises the second plurality of areas of interest ZR2.

[0062] In examples, prior to the operation 160 of determining the detection index, the method 100 may comprise an operation 155 of determining a region confidence index associated with a region of interest R. The region confidence index associated with a region of interest R is determined from at least one zone confidence index of a zone of interest Z r belonging to the region of interest R.

[0063] In examples, a confidence index of a region of interest R is associated with a specific color, such that a specific region of interest is associated with a plurality of confidence indices. A confidence index of a region of interest R associated with a specific color is determined from at least one area confidence index associated with that specific color and associated with an area of ​​interest Z r belonging to the region of interest R.

[0064] In the examples comprising the operation 155 of determining a region confidence index, the priority vehicle detection index of the detected vehicle is determined from at least one region confidence index associated with at least one region of interest R of the detected vehicle. Advantageously, the priority vehicle detection index for a detected vehicle can be determined from a plurality of region confidence indices associated with at least one region of interest R determined for this vehicle.

[0065] In particular, in examples, the priority vehicle detection index for a specific vehicle may be determined from the region confidence index having the highest value among the indices associated with the regions of interest R determined for the specific vehicle. The other region confidence indices are therefore not retained in the determination of the priority vehicle detection index so as to reduce the number of mathematical operations performed to determine this index, which reduces the complexity of the method while allowing faster execution. Furthermore, priority vehicles rarely have emergency lights in two different regions of interest R so that taking into account the confidence indices calculated in several regions of interest R to determine the detection index is not always relevant.

[0066] It is understood here that to the extent that the camera on board the vehicle acquires images in real time, the confidence indices associated with the areas of interest Z r and possibly to the regions of interest R and the priority vehicle detection index of a vehicle detected by the method 100 can be updated with each new image acquired. The detected vehicles are therefore tracked with a modification of the different indices as the images are acquired in order to update the priority vehicle detection index in real time.

[0067] In examples, the method 100 may also comprise an operation 152 of classifying a region of interest Z r as active or inactive based on an area confidence index associated with that area of ​​interest Z r. In particular, the area confidence index can be compared to an activation threshold so that the area of ​​interest is classified as active when the area confidence index is greater than the activation threshold. The activation threshold used can correspond to a hysteresis threshold. Furthermore, it is understood that an area classified as active can then be classified as inactive when the area confidence index is recalculated upon acquisition of a new image by the on-board camera.

[0068] In examples in which an area confidence index of an area of ​​interest Z r is associated with a specific color, the area of ​​interest Z rcan be classified as active or inactive for the specific color. This classification is done based on the area confidence index associated with the specific color. The area of ​​interest is therefore classified as active for the specific color when the confidence index associated with that color is greater than the activation threshold.

[0069] In examples, a region confidence index of a region of interest R may be determined from a weight obtained as a function of: - a positioning of the areas of interest Z r classified as active in the region of interest; and / or - a color associated with the areas of interest classified as Z r as active in the region of interest.

[0070] In examples, the weight may scale proportionally with the number of active areas of interest in the region of interest R.

[0071] In examples, the weight may scale proportionally to the number of different colors for which the areas of interest in a region are active in the region of interest R.

[0072] In examples, when multiple areas of interest are active for different colors, the weight obtained is greater than when these areas are active for a single color.

[0073] In examples, a weight may be assigned to patterns formed by a selection of areas of interest Z r so that when the areas of interest Z rforming a pattern are classified as active, the weight used to determine the confidence index of the region R corresponds to the weight assigned to the pattern considered. The pattern can be a geographical pattern (arrangement of active areas of interest in the region of interest R) and / or a colorimetric pattern (one or more active areas of interest for different colors). A pattern is therefore understood to mean a selection of areas of interest Z r classified as active or non-active, in other words, an arrangement of flashing emergency lights.

[0074] An example of a pattern to which a weight can be associated may for example correspond to a pattern comprising a first area of ​​interest classified as active and a second area of ​​interest classified as active, the first and second areas of interest being separated from each other by at least one area classified as inactive. Several examples of this type of pattern are shown in Figure 4 illustrating the areas of interest ZRI associated with the first region of interest R1v. In the examples illustrated in Figure 4, the filled ZRI areas of interest are considered to be the areas classified as active while the empty ZRI areas of interest are considered to be the areas classified as inactive.

[0075] In a particular example, a region confidence index of a specific region of interest R may correspond to the maximum value of the area confidence index of the regions of interest Z rincluded or composing the specific region of interest R multiplied by the weight obtained from the positioning and / or from the color of the areas of interest Z r of the specific R region of interest classified as active.

[0076] In examples in which the areas of interest Z r can be activated for different colors, operation 161 of identifying the vehicle detected as corresponding to a priority vehicle can be carried out: - when the priority vehicle detection index is higher than a predetermined detection threshold, and - when a confidence index associated with a specific color corresponding to a color used by priority vehicles in the geographical area in which the specific vehicle carrying the camera is circulating is higher than a predetermined threshold.

[0077] Indeed, the color of the flashing emergency lights of priority vehicles depends in particular on the legislation of the geographical area in which the vehicle is traveling, this legislation being able to be different depending on the country. Therefore, taking into account the color of the lights selected as being able to correspond to a flashing emergency light in the identification of a priority vehicle makes it possible to adapt to the different colors used by priority vehicles in the different geographical areas of the world. This thus makes it possible to avoid erroneous identifications when the colors identified on the detected vehicle and the colors used by the priority vehicles in the geographical area in which the specific vehicle carrying the camera is traveling do not match.

[0078] Thus, the method presented by the present disclosure makes it possible to determine an index indicating a probability that a vehicle detected by a camera on a vehicle is a priority vehicle, and in certain examples, makes it possible to determine whether or not the detected vehicle is a priority vehicle. The method is therefore particularly advantageous insofar as driving assistance functions can use this information to act accordingly, and in particular to facilitate the passage of priority vehicles. detected where appropriate. Furthermore, since the method 100 relies on pre-existing vehicle detection and light detection functions, its implementation does not require significant modification of the consequent software architecture present in vehicles using driving assistance functions and / or autonomous driving functions.

Claims

Claims

1. A method of processing a plurality of images acquired by a camera mounted on a specific vehicle, each image of the plurality of images being associated with a respective acquisition time, the method comprising the following operations: - detecting (110) at least one vehicle traveling in the same direction as the specific vehicle from the plurality of acquired images; - determining (120), for a detected vehicle, a plurality of areas of interest (Z r ) of the vehicle shared by images of the plurality of images and being capable of comprising at least one flashing emergency light of a priority vehicle; - detecting (130) the lights present on the images of the plurality of images; - construct (140) a history indicating, for each image of the plurality of images, a presence or an absence of a light in each of the zones of interest (Z r) from the detected lights; - determine (150), for at least one area of ​​interest (Z r ), an area confidence index from the information associated with the area of ​​interest (Z r ) concerned listed in the history; and - determining (160) a priority vehicle detection index indicating a probability that the detected vehicle is a priority vehicle from a zone confidence index of at least one zone of interest (Z r ) of the detected vehicle.

2. Method according to the preceding claim, in which, the determination (160) of a zone confidence index of an area of ​​interest (Z r ) includes, when a light of the same color has been detected on several images of the plurality of images as included at least in part in the area of ​​interest (Z r ), a determination (151) of at least one characteristic among: - a detected light flashing frequency; - a variance in flickering of the detected light; - an average duration of the phases during which the detected light is on; or - an average duration of the phases during which the detected light is off; and the confidence index associated with an area of ​​interest (Z r ) of the detected vehicle is determined from at least one characteristic.

3. Method according to one of the preceding claims, in which an area confidence index is associated with a specific color, so that an area of ​​interest (Z r ) specific is associated with a plurality of confidence indices when a plurality of lights of different colors were respectively detected on several images of the plurality of images as present in the area of ​​interest (Z r ).

4. A method according to any preceding claim, wherein the method further comprises the following steps: - determine, for a detected vehicle, at least one region of interest (R) comprising several areas of interest; - determining (155), for each region of interest, a region confidence index, the region confidence index of a region of interest being determined from at least one area confidence index of an area of ​​interest belonging to the region of interest; and wherein a priority vehicle detection index is determined from a region confidence index of at least one region of interest.

5. A method according to the preceding claim, wherein the method further comprises the following operation: - classify (152) an area of ​​interest (Z r) as being active or inactive depending on a value of a confidence index associated with it; and a region confidence index of a region of interest (R) is also determined from a weight calculated depending on at least one of: a) a positioning of the areas of interest (Z r ) classified as active in a region of interest; or b) a color associated with the areas of interest (Z r ) classified as active in a region of interest.

6. A method according to any one of claims 4 or 5, wherein a specific detected vehicle is a car or a truck; and the at least one specific region of interest (R) determined for the specific detected vehicle comprises at least one region of interest (R) from: a) a first region of interest (R1v) corresponding to an upper portion of the specific vehicle positioned at roof level in the image; or b) a second region of interest (R2v) corresponding to a central strip of the specific vehicle and extending between the two front lights of the specific vehicle.

7. Method according to the preceding claim, in which the second region of interest (R2v) is devoid of the vehicle's turn signals.

8. A method according to any one of claims 1 to 5, wherein a specific detected vehicle is a motorcycle and the plurality of areas of interest (Z r) of the motorcycle includes at least one area of ​​interest (Z r ) among: a) a first zone of interest (ZRmi) corresponding to a first sideband extending from a first end of the motorcycle windshield and comprising a first rearview mirror of the motorcycle; b) a second area of ​​interest (Z Rm 2) extending between two ends of the motorcycle windshield located on the same horizontal plane; or c) a third area of ​​interest (Z Rm 3) corresponding to a second side strip extending from a second end of the windshield of the motorcycle and comprising a second rearview mirror of the motorcycle.

9. A method according to any preceding claim, wherein, when the priority vehicle detection index is greater than a predetermined detection threshold for a detected vehicle, the method further comprises the following step: - identify (161) the detected vehicle as corresponding to a priority vehicle.

10. A data processing device (10) configured to implement any one of the methods according to claims 1 to 9.

Images