Seat Belt Wearing Detection Method, Device, Storage Medium, and Processor

Abstract

The present application discloses a seat belt fastening detection method, device, storage medium and processor, which includes: acquiring a target image collected from a traffic device; identifying a first image region and / or a second image region in the target image, where the first image region is a region in the target image where the head and trunk of a target object are located, and the second image region is a region in the target image where the face of the target object is located; identifying a third image region based on the first image region and / or the second image region, where the third image region is a seat belt fastening region in the target image; and determining whether the fastening of a seat belt in the target image satisfies a predetermined fastening requirement based on image information in the third image region.

Description

Technical Field

[0001] This application claims the priority of a Chinese patent application with an application number of 202011570898.8 and an application title of "Seat Belt Wearing Detection Method, Device, Storage Medium and Processor", which was filed with the China Patent Office on December 26, 2020, and all of its contents are incorporated herein by reference.

[0002] This application relates to the technical field of image recognition, and specifically to a seat belt wearing detection method, device, storage medium and processor.

Background Art

[0003] With the development of cities and the improvement of residents' living standards, the requirements for various vehicles are increasing more and more. Accordingly, technological innovation and industrial upgrading are being carried out around how to ensure the safety of the personnel inside the vehicle during driving.

[0004] As a close protection means when the personnel inside the vehicle encounter an emergency during driving, in some cities, there are also requirements for passengers in the passenger seat of commercial vehicles to be forced to wear seat belts. Therefore, a method for accurately determining whether the user is wearing a seat belt has also attracted more attention.

[0005] In the related art, the solution for seat belt discrimination mainly relies on whether the buckle at the tip of the seat belt of the passenger or driver is hooked in the detection groove. Specifically, after connecting the tip of the seat belt and the detection groove, a detection circuit is formed inside the vehicle, and the system obtains the feedback that the seat belt has been worn. However, this judgment mechanism not only increases the complexity of the physical devices and the system inside the vehicle and raises the cost, but also has vulnerabilities in the detection mechanism, making it easy to be exploited by those who avoid wearing the seat belt, and it is impossible to avoid deceptive means adopted because the relevant person intentionally does not wear it (for example, by inserting a locking buckle similar to the tip of the seat belt into the detection groove, the system is misrecognized that the passenger at that position has already worn the seat belt). Also, for example, it is impossible to more deeply judge the state such as whether the occupant is wearing the seat belt correctly.

[0006] Regarding the problem that it is difficult to accurately determine whether a passenger is wearing a seat belt correctly in the related art, currently, no effective solution technical means have been proposed.

Summary of the Invention

Problems to be Solved by the Invention

[0007] In order to solve the problem that it is difficult to accurately determine whether a passenger is wearing a seat belt correctly in the related art, at least some embodiments of the present invention provide a seat belt wearing detection method, device, storage medium, and processor.

Means for Solving the Problems

[0008] According to an embodiment of the present invention, a seat belt wearing detection method is provided. The method includes: obtaining a target image collected from a transportation device; identifying a first image region and / or a second image region in the target image, where the first image region is a region where the head and torso of a target object are located in the target image, and the second image region is a region where the face of the target object is located in the target image; identifying a third image region based on the first image region and / or the second image region, where the third image region is a seat belt wearing region in the target image; and determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region.

[0009] Preferably, when the upper body region of the target object appears in the target image, the region where the head and torso of the target object are located is the smallest region including the face, neck, shoulders, and upper chest, and the region where the face of the target object is located is the smallest region including the face.

[0010] Preferably, before obtaining the target image collected from the transportation device, the method further includes: obtaining a plurality of images collected from the transportation device, marking whether the obtained images include a head and torso region, and marking the position of the head and torso region for the images including the head and torso region to obtain a plurality of first marked images; identifying the plurality of first marked images as first training set data, training a neural network model with the first training set data to obtain a first detector configured to detect the head and torso region; and / or obtaining a plurality of images collected from the transportation device, marking whether the obtained images include a face region, and marking the position of the face region for the images including the face region to obtain a plurality of second marked images; identifying the plurality of second marked images as second training set data, training a neural network model with the second training set data to obtain a second detector configured to detect the face region.

[0011] Preferably, when obtaining the first detector and the second detector, specifying the first image region in the target image includes detecting the target image by the first detector and obtaining the first image region; specifying the second image region in the target image includes detecting the target image by the second detector and obtaining the second image region; when obtaining the first detector, specifying the first image region in the target image includes detecting the target image by the first detector and obtaining the first image region; specifying the second image region in the target image includes cutting out the second image region from the first image region based on the predetermined sitting posture information of the target object; when obtaining the second detector, specifying the second image region in the target image includes detecting the target image by the second detector and obtaining the second image region; specifying the first image region in the target image includes expanding the second image region in the target image downward, leftward, and rightward by a predetermined region based on the predetermined sitting posture information of the target object and obtaining the first image region.

[0012] Preferably, specifying the third image region based on the first image region and / or the second image region includes, when obtaining the first image region in the target image and obtaining the second image region in the target image, specifying the first seat belt wearing region in the target image based on the offset distance of the covering region with respect to the first image region when the seat belt is correctly worn, and specifying the second seat belt wearing region in the target image based on the offset distance of the covering region with respect to the second image region when the seat belt is correctly worn, and averaging the first seat belt wearing region and the second seat belt wearing region at the spatial position of the target image to obtain the third image region.

[0013] Preferably, before specifying the third image region based on the first image region and / or the second image region, the method further includes acquiring a plurality of head and torso region images, marking whether the acquired images include a seat belt wearing region, and marking the position of the facial region for the images including the seat belt wearing region to obtain a plurality of third marked images; specifying the plurality of third marked images as third training set data, training a neural network model with the third training set data, and obtaining a third detector configured to detect the seat belt wearing region. When the first image region in the target image and the second image region in the target image are obtained, specifying the third image region based on the first image region and / or the second image region includes inputting the position of the first image region in the target image, the position of the second image region in the target image, and the first image region into the third detector to detect the third image region.

[0014] Preferably, before specifying the third image region based on the first image region and / or the second image region, the method further includes acquiring a plurality of head and torso region images, marking whether the acquired images include a seat belt wearing region, and marking the position of the seat belt wearing region for the images including the seat belt wearing region to obtain a plurality of fourth marked images; specifying the plurality of fourth marked images as fourth training set data, training a neural network model with the fourth training set data, and obtaining a fourth detector configured to detect the seat belt wearing region. When the first image region is detected in the target image, specifying the third image region based on the first image region and / or the second image region includes detecting the first image region by the fourth detector and obtaining the third image region.

[0015] Preferably, before identifying the third image region based on the first image region and / or the second image region, the method further includes obtaining a plurality of images including a facial region, marking whether the obtained images include a seat belt wearing region, and marking the position of the seat belt wearing region for the images including the seat belt wearing region, to obtain a plurality of fifth marked images; identifying the plurality of fifth marked images as fifth training set data, training a neural network model with the fifth training set data, and obtaining a fifth detector configured to detect the seat belt wearing region. When the second image region is detected in the target image, identifying the third image region based on the first image region and / or the second image region includes detecting the second image region by the fifth detector and obtaining the third image region.

[0016] Preferably, before determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region, the method further includes obtaining a plurality of seat belt wearing region images, marking whether the wearing of the seat belt in the obtained images meets the predetermined wearing requirement, to obtain a plurality of sixth marked images; identifying the plurality of sixth marked images as sixth training set data, training a neural network model with the sixth training set data, and obtaining a discriminator configured to determine whether the wearing of the seat belt in the target image meets the predetermined wearing requirement. Determining whether the wearing of the seat belt in the target image meets the predetermined wearing requirement based on the image information in the third image region includes detecting the third image region by the discriminator and obtaining a conclusion on whether the wearing of the seat belt meets the predetermined wearing requirement.

[0017] Preferably, after determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area, if the wearing of the seat belt does not meet the predetermined wearing requirement, the method further includes issuing prompt information configured to prompt the passenger to correctly wear the seat belt.

[0018] Another embodiment of the present invention further provides a seat belt wearing detection device. The device includes a first acquisition unit configured to acquire a target image collected from a transportation device, and a first identification unit configured to identify a first image area and / or a second image area in the target image, where the first image area is an area where the head and torso of the target object in the target image are located, and the second image area is an area where the face of the target object in the target image is located. The device further includes a second identification unit configured to identify a third image area based on the first image area and / or the second image area, where the third image area is a seat belt wearing area in the target image, and a determination unit configured to determine whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area.

[0019] Another embodiment of the present invention further provides a non-volatile storage medium including a stored program, where when the program is executed, the non-volatile storage medium controls a device in which the non-volatile storage medium is arranged to execute a seat belt wearing detection method.

[0020] Another embodiment of the present invention provides an electronic device including a processor and a memory, where computer-readable instructions are stored in the memory, and the processor is configured to execute the computer-readable instructions. When the computer-readable instructions are executed, the electronic device executes a seat belt wearing detection method.

[0021] According to the present application, there are steps of obtaining a target image collected from a transportation device, and identifying a first image region and / or a second image region in the target image, where the first image region is a region where the head and torso of a target object are located in the target image, and the second image region is a region where the face of the target object is located in the target image; identifying a third image region based on the first image region and / or the second image region, where the third image region is a seat belt wearing region in the target image; and determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region. By adopting these steps, the problem that it is difficult to accurately determine whether a passenger is wearing a seat belt correctly in the related art is solved. By identifying the seat belt wearing region based on the region where the head and torso of the target object and / or the face of the target object are located in the target image, it is determined whether the wearing of the seat belt meets a predetermined wearing requirement, and further the effect of improving the accuracy of determining whether a passenger is wearing a seat belt correctly is achieved.

[0022] The drawings constituting a part of the present application are provided for further understanding of the present application. The exemplary embodiments of the present application and their descriptions are for interpreting the present application and do not unduly limit the present application. Each drawing is as follows.

Brief Description of the Drawings

[0023]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Embodiments for Carrying out the Invention

[0024] Unless there is a contradiction, the embodiments and features in the embodiments in the present application can be combined with each other. Hereinafter, the present application will be described in detail with reference to the drawings by giving examples.

[0025] To enable those skilled in the art to better understand the solution means of the present application, the technical solution means in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. It is obvious that the described embodiments are only some of the embodiments of the present application, not all of them. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the present application.

[0026] Note that the terms "first", "second", etc. in the description, claims, and the above drawings of this application are for distinguishing similar objects and not for explaining a specific order or sequence. The data used in this way is for easily explaining the embodiments of this application here, and it should be understood that it can be exchanged in appropriate situations. Also, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, and may include other steps or units that are not clearly listed or are specific to those processes, methods, products, or devices.

[0027] Example 1 According to Example 1 of this application, a seat belt wearing detection method is provided.

[0028] FIG. 1 is a flowchart of the seat belt wearing detection method according to Example 1 of this application. As shown in FIG. 1, the method includes the following steps.

[0029] Step S102: Obtain a target image collected from a transportation device.

[0030] Specifically, the target image collected from the transportation device may be a scene image in the driver's cab or passenger compartment of the transportation device. Here, the transportation device may be a vehicle, or may also be a device such as a cabin or ferry where passengers need to wear seat belts. Example 1 of this application does not limit the type of transportation device.

[0031] Step S104: Identify a first image area and / or a second image area in the target image. The first image area is the area where the head and torso of the target object in the target image are located, and the second image area is the area where the face of the target object in the target image is located.

[0032] Specifically, the target object may be a human body. The first image region refers to the region of interest (ROI) of the head and torso identified when detecting the head and torso of the human body. Since the English for head and torso detection is Human Torso Detection and it is abbreviated as HTD, the first image region can be abbreviated as the HTROI region. The second image region refers to the region of interest of the human face identified when detecting the human face. Since the English for face detection is Face Detection and it is abbreviated as FTD, the second image region can be abbreviated as FTROI.

[0033] Here, the region of interest refers to the rectangular sub-image of a certain region in the complete image. Taking the HTROI region as an example, assuming that the resolution of the complete target image is 1280×800, and the coordinates of the upper left corner of the regressed HTROI region are (100, 100) and the coordinates of the lower right corner are (350, 780), the HTROI region in the target image only includes the content within the region from (100, 100) to (350, 780) in the complete image, which is equivalent to cutting out a region with a resolution of 250×680 from (100, 100) in the target image.

[0034] Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, when the target image includes the upper body region of the target object, the region where the head and torso of the target object are located is the smallest region including the face, neck, shoulders, and upper chest, and the region where the face of the target object is located is the smallest region including the face.

[0035] For example, when the target object is a driver sitting in the driver's seat, assuming that the entire image includes the upper body of the driver sitting in the driver's seat, the first image region is specifically defined as the smallest region that completely includes the human face, neck, shoulders, and upper chest. As shown in Figure 2, the region surrounded by the square frame is the first image region. The second image region is specifically defined as the smallest region that completely includes the human face on the premise that the upper body of the driver sitting in the driver's seat appears entirely in the entire image.

[0036] In step S106, based on the first image region and / or the second image region, a third image region is identified, and the third image region is the seat belt wearing region in the target image.

[0037] Specifically, the third image region refers to the seat belt wearing interest region identified when detecting the seat belt wearing region. Since the English of seat belt is Safety Belt, the third image region can be abbreviated as SROI.

[0038] In the first embodiment of the present application, based on the size of the seat belt in the actual scene, by detecting the face and the head and torso, the detection region of the seat belt is reduced, the features of the seat belt itself are made prominent, and the recognizable degree of the seat belt wearing region can be improved.

[0039] In step S108, based on the image information in the third image region, it is determined whether the wearing of the seat belt in the target image meets a predetermined wearing requirement.

[0040] Specifically, after obtaining the third image region, by comparing the information such as the actual wearing position of the seat belt in the third image region with the information such as the position where the seat belt is correctly worn, a conclusion can be obtained as to whether the seat belt is correctly worn. Further, by training a discriminator configured to determine whether the seat belt is correctly worn based on a plurality of images including the seat belt wearing region, the discriminator can determine whether the seat belt is correctly worn.

[0041] According to the first embodiment of the present application, a non-contact recognition of the seat belt wearing state of personnel based on vision is realized, the production cost of the device is saved, and at the same time, the result of determining the seat belt wearing state of personnel becomes more accurate.

[0042] The seat belt wearing detection method provided by Example 1 of this application includes obtaining a target image collected from a transportation device, and identifying a first image region and / or a second image region in the target image, where the first image region is the region in the target image where the head and torso of the target object are located, and the second image region is the region in the target image where the face of the target object is located, and identifying a third image region based on the first image region and / or the second image region, where the third image region is the seat belt wearing region in the target image, and determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region, thereby solving the problem that it is difficult to accurately determine whether a passenger is wearing a seat belt correctly in the related art. By identifying the seat belt wearing region based on the region where the head and torso of the target object are located and / or the region where the face is located in the target image, it is determined whether the wearing of the seat belt meets a predetermined wearing requirement, and further the effect of improving the accuracy of determining whether a passenger is wearing a seat belt correctly is achieved.

[0043] Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, before obtaining a target image collected from a transportation device, the method includes obtaining a plurality of images collected from the transportation device, marking whether the obtained images include a head and torso region, and marking the position of the head and torso region for the images including the head and torso region, to obtain a plurality of first marked images; identifying the plurality of first marked images as first training set data, and training a neural network model with the first training set data to obtain a first detector configured to detect the head and torso region; and / or obtaining a plurality of images collected from the transportation device, marking whether the obtained images include a face region, and marking the position of the face region for the images including the face region, to obtain a plurality of second marked images; identifying the plurality of second marked images as second training set data, and training a neural network model with the second training set data to obtain a second detector configured to detect the face region.

[0044] Note that the training set of the detector based on the convolutional neural network (CNN) in Embodiment 1 of the present application can be obtained in the following manner. When the target object is a driver sitting at the driver's seat in the vehicle, first, a plurality of complete images of the scene at the driver's seat position in the vehicle can be obtained by a general data acquisition method (for example, actual sampling). Specifically, the amount of image data includes at least 10,000 images. Next, qualitative marking of whether each image contains the target content is performed one by one, and then quantitative marking of the ROI position of the target content is continuously performed for the images containing the content. For each quantitative marking of the ROI position, it includes the two-dimensional coordinate pair of the upper left and lower right of the target content in the complete image and the presence or absence of target inclusion. All the images and related qualitative and quantitative marking contents constitute the training set of the detector.

[0045] Furthermore, after constructing the neural network, the image data in the training set is randomly selected according to different batches and then input into the neural network. Weighting operations are performed using different weights in the network. Various qualitative and quantitative conclusions and corresponding marking data included in the output result of the neural network are collated with the input data, and a detector based on CNN can be trained to adjust the weight values in the neural network based on the collation error. The entire process is regarded as one training. After multiple trainings, the output conclusions of the neural network are gradually matching the actual calibration conclusions with the adjustment of the weights, and when the difference between the two is smaller than a certain degree or the number of training times reaches a sufficient number, the training is stopped, and the parameters of the detector are specified based on the weights in the neural network obtained by training.

[0046] Specifically, when training the first detector (HTROI region detector) configured to detect the head and torso region when the target object is a driver in the driver's seat position, the target content of the training set is the HTROI region, the input data is the image of the entire in-vehicle scene, the quantitative marking content is the coordinate pair of the upper left and lower right of the HTROI region, and the qualitative marking content is the presence or absence of inclusion in the HTROI region. When training the second detector (FROI region detector) configured to detect the face region, the target content of the training set is the FROI region, the input data is the image of the entire in-vehicle scene, the quantitative marking content is the coordinate pair of the upper left and lower right of the FROI region, and the qualitative marking content is the presence or absence of inclusion in the FROI region.

[0047] According to Embodiment 1 of the present application, both the HTROI region detector and the FROI region detector are realized in the manner of CNN forward inference. Compared with the conventional image recognition method, the recognition effect is better, it can effectively adapt to the optimization method of artificial intelligence hardware, and the detection performance is better.

[0048] When obtaining the first detector and / or the second detector, there are multiple methods for identifying the first image region. Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, when obtaining the first detector and the second detector, identifying the first image region in the target image includes detecting the target image with the first detector to obtain the first image region, and identifying the second image region in the target image includes detecting the target image with the second detector to obtain the second image region.

[0049] When obtaining the first detector, identifying the first image region in the target image includes detecting the target image with the first detector to obtain the first image region, and identifying the second image region in the target image includes cropping the second image region from the first image region based on the predetermined sitting posture information of the target object.

[0050] When obtaining the second detector, identifying the second image region in the target image includes detecting the target image with the second detector to obtain the second image region, and identifying the first image region in the target image includes expanding the second image region downward, leftward, and rightward by a predetermined region in the target image based on the predetermined sitting posture information of the target object to obtain the first image region.

[0051] In addition, in one selectable embodiment, by obtaining the HTROI region detector and the FROI region detector respectively by a method based on CNN, the two-dimensional positions of the HTROI region and the FROI region in the target image based on the input of the entire image can be directly regressed.

[0052] In one selectable embodiment, first obtain the HTROI region detector by the method of CNN, detect the HTROI region with the HTROI region detector, and then based on the approximate predicted posture when the human body is sitting at the driver's seat position, the FROI region can also be obtained at the center of the upper half of the HTROI region.

[0053] In one selectable embodiment, first, an FROI region detector is obtained by means of the CNN method, the FROI region is detected by the FROI region detector, and then, based on the approximate predicted posture when the human body is seated at the driver's seat position, the FROI region can be expanded downward and on both sides to obtain an HTROI region.

[0054] When the first image region and / or the second image region are obtained, there are multiple methods for specifying the third image region. Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, based on the first image region and / or the second image region, specifying the third image region includes: when the first image region in the target image is obtained and the second image region in the target image is obtained, based on the offset distance of the covering region with respect to the first image region when the seat belt is correctly worn, specifying the first seat belt wearing region in the target image, and based on the offset distance of the covering region with respect to the second image region when the seat belt is correctly worn, specifying the second seat belt wearing region in the target image, and averaging the first seat belt wearing region and the second seat belt wearing region at the spatial position of the target image to obtain the third image region.

[0055] Specifically, when both the HTROI region and the FROI region are detected in the target image, and when the seat belt is correctly worn, two SROI regions for detecting whether the human body is wearing the seat belt in the entire image are set based on the respective offsets of the covering region with respect to the HTROI region and the FROI region. After averaging these two SROI regions at the spatial position, the final SROI region can be obtained, thereby building a data basis for detecting whether the seat belt is correctly worn.

[0056] Preferably, in the seat belt wearing detection method provided by Example 1 of the present application, before identifying the third image region based on the first image region and / or the second image region, the method further includes acquiring a plurality of head and torso region images, marking whether the acquired images include a seat belt wearing region, and marking the position of the face region for the images including the seat belt wearing region, to obtain a plurality of third marked images; identifying the plurality of third marked images as third training set data, training a neural network model with the third training set data, and obtaining a third detector configured to detect the seat belt wearing region. When obtaining the first image region and / or the second image region in the target image, identifying the third image region includes identifying the position of the first image region in the target image and / or the position of the second image region in the target image, and inputting the first image region into the third detector to detect the third image region.

[0057] Specifically, when training a third detector configured to detect the seat belt wearing region where the target object is a driver sitting in the driver's seat, the target content of the training set is SROI, the input data is the image range limited by the HTROI region, and the quantitative marking content is the coordinate pair of the upper left and lower right of the FROI region in the image range coordinate system limited by the HTROI region.

[0058] Furthermore, after obtaining the third detector, by directly inputting the position coordinates of the HTROI region and the FROI region in the image (the two-dimensional coordinate values of the two points of the upper left and lower right of the HTROI region, the two-dimensional coordinate values of the two points of the upper left and lower right of the FDROI, a total of four two-dimensional coordinate values of the points) and the image of the HTROI region in the whole image into the pre-trained third detector, the final SROI region can be regressed.

[0059] There are multiple methods for identifying the third image region. Preferably, in the seat belt wearing detection method provided by Example 1 of this application, based on the first image region and / or the second image region, before identifying the third image region, the method includes acquiring a plurality of head and torso region images, marking whether the acquired images include a seat belt wearing region, and marking the position of the seat belt wearing region for the images including the seat belt wearing region to obtain a plurality of fourth marked images; identifying the plurality of fourth marked images as fourth training set data, and training a neural network model with the fourth training set data to obtain a fourth detector configured to detect the seat belt wearing region. When the first image region is detected in the target image, identifying the third image region based on the first image region and / or the second image region includes detecting the first image region by the fourth detector and obtaining the third image region.

[0060] Specifically, a fourth detector configured to detect the seat belt wearing region can be trained with a plurality of head and torso region images. The target content of the training set is the STROI region, the input data is the HTROI region image, the quantitative marking content is the coordinate pair of the upper left and lower right of the SROI region, and the qualitative marking content is the presence or absence of inclusion of the SROI region. Further, after training to obtain the fourth detector, the HTROI region is detected by the fourth detector to obtain the STROI region.

[0061] Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, before identifying the third image area based on the first image area and / or the second image area, the method further includes acquiring a plurality of images including a face area, marking whether the acquired images include a seat belt wearing area, and marking the position of the seat belt wearing area for the images including the seat belt wearing area, to obtain a plurality of fifth marked images; identifying the plurality of fifth marked images as fifth training set data, and training a neural network model with the fifth training set data to obtain a fifth detector configured to detect the seat belt wearing area. When the second image area is detected in the target image, identifying the third image area based on the first image area and / or the second image area includes detecting the second image area with the fifth detector and obtaining the third image area.

[0062] Specifically, a fifth detector configured to detect a seat belt wearing area with a plurality of images including a face area can be trained. The target content of the training set is the STROI area, the input data is the FROI area image, the quantitative marking content is the coordinate pair of the upper left and lower right of the SROI area, and the qualitative marking content is the presence or absence of inclusion in the SROI area. Further, after training to obtain the fifth detector, the FROI area is detected with the fifth detector to obtain the STROI area.

[0063] Also, when neither the HTROI area nor the FROI area is detected in the target image, it is considered that there is no occupant in the vehicle at that time, and a determination of non-wearing of the seat belt (because there is no person) is directly given at that time.

[0064] After obtaining the third-region image, based on the third-region image, determine the wearing state of the seat belt. Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, before determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region, the method includes acquiring a plurality of seat belt wearing region images, marking whether the wearing of the seat belt in the acquired images meets a predetermined wearing requirement, and obtaining a plurality of sixth-marked images; specifying the plurality of sixth-marked images as sixth-training set data, training a neural network model with the sixth-training set data, and obtaining a discriminator configured to determine whether the wearing of the seat belt in the target image meets a predetermined wearing requirement. Determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region includes detecting the third image region by the discriminator and obtaining a conclusion on whether the wearing of the seat belt meets a predetermined wearing requirement.

[0065] Specifically, based on the CNN, a discriminator configured to determine whether the wearing of the seat belt meets a predetermined wearing requirement can be trained. The target content of its training set is a qualitative conclusion on whether the seat belt is worn correctly. The input data is the image range limited by the SROI, and the qualitative marking content is whether the seat belt is worn correctly.

[0066] Furthermore, after obtaining the discriminator, input the image data of the SROI part in the image data of the entire scene into the discriminator based on the pre-trained CNN to determine whether the occupant is wearing a seat belt.

[0067] For example, as shown in FIG. 3, the area surrounded by the black frame is the detected HTROI area, the area surrounded by the white frame is the detected FROI area, and the area surrounded by the gray frame is the SROI area obtained from the FROI area and the HTROI area. After inputting the data of the SROI area into the seat belt discriminator, an accurate conclusion of "With Safety Belt" can be obtained.

[0068] When the solution means according to the embodiment of the present application is not adopted, for example, when the entire target image is input into the seat belt discriminator, as shown in FIG. 4, the seat belt feature is weaker in the entire image compared to the features of a person's face or body, and it is difficult for the seat belt discriminator to recognize. Therefore, in the same scene, it is difficult to determine the SROI area, and an incorrect conclusion of "Without Safety Belt" is obtained.

[0069] Preferably, in the seat belt wearing detection method provided by Embodiment 1 of the present application, after determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area, the method further includes issuing prompt information for prompting the passenger to correctly wear the seat belt when the wearing of the seat belt does not meet the predetermined wearing requirement.

[0070] Specifically, when the seat belt is worn incorrectly, information indicating that the occupant is not wearing the seat belt correctly or is not wearing the seat belt at all is issued to prompt the occupant to wear the seat belt correctly.

[0071] FIG. 5 is a flowchart of a selectable seat belt wearing detection method according to Embodiment 1 of the present application.

[0072] The method according to Embodiment 1 of the present application can be executed on a PC, a mobile phone, and currently mainstream in-vehicle embedded platforms, and the in-vehicle embedded platforms include, but are not limited to, an ARM architecture, a DSP architecture, and an ARM+DSP architecture.

[0073] As shown in FIG. 5, in this method, in-vehicle scene image data is acquired, human torso detection (HTD) is performed on the entire in-vehicle scene image data, and a human torso region of interest (Human Torso ROI, HTROI) is acquired. At the same time, face detection (FD) is performed on the entire in-vehicle scene image data, and a face region of interest (Face ROI, FROI) is acquired.

[0074] Furthermore, based on the positional relationship between the HTROI region and the FROI region, a seatbelt ROI (SROI) is estimated, and the positional coordinates of the HTROI region and the FROI region in the image and the data of the HTROI region part in the image data of the entire scene are directly input into a regressor based on a pre-trained convolutional neural network CNN to regress the SROI region.

[0075] In addition, when only one of HTD and FD detects an ROI region, the SROI region may be estimated based on the positional relationship using only the detected ROI. When neither the HTROI region nor the FROI region is detected, it is considered that there is no occupant in the vehicle at this time, and at this time (because there is no person), a judgment that the seatbelt is not worn is directly given.

[0076] Finally, the image data of the SROI region in the entire scene image data is input into a seatbelt discriminator based on a pre-trained convolutional neural network CNN to determine whether the occupant is wearing a seatbelt. If the seatbelt is worn incorrectly, it is also presented that the occupant is not wearing a seatbelt.

[0077] According to Example 1 of the present application, non-contact recognition of the seat belt wearing state of personnel based on vision is realized, saving the production cost of the device, and at the same time, the result of judging the seat belt wearing state of personnel becomes more accurate. On the other hand, based on the size of the seat belt in the actual scene, by facial detection and head and torso detection, the detection area of the seat belt is reduced, the characteristics of the seat belt itself are made prominent, and the recognizable degree of the seat belt wearing area is improved.

[0078] Figure 6 is a schematic diagram of a selectable seat belt wearing detection method according to Example 1 of the present application.

[0079] The method according to Example 1 of the present application can be executed on a PC, a mobile phone, and the current mainstream in-vehicle embedded platform. The in-vehicle embedded platform includes, but is not limited to, the ARM architecture, the DSP architecture, and the ARM+DSP architecture.

[0080] As shown in Figure 6, in this method, when the occupant sits on the seat in a normal posture and faces forward, a plurality of frames of images are collected from the entire in-vehicle scene data obtained, and the region of interest (Face ROI, FROI) of the human face is obtained. Since the posture of the occupant is normal, the corresponding FROI region can always be obtained from each frame of the image.

[0081] Furthermore, averaging is performed based on the positions of the FROI regions of the plurality of frames of images to obtain an averaged FROI region. Further, based on the averaged FROI region, the seat belt wearing region of interest (Seatbelt ROI (Region of Interest), SROI) is obtained by expanding based on the positional relationship between the predicted FROI region and the SROI.

[0082] Furthermore, for the entire in-vehicle scene image input for each subsequent frame, the image data of the SROI portion is input into a seat belt discriminator based on a pre-trained CNN to determine whether the occupant is wearing a seat belt. Even if the seat belt is worn incorrectly, it is presented that the occupant is not wearing a seat belt. Specifically, as shown in FIG. 7, the area surrounded by the white frame is the detected FROI area, and the area surrounded by the gray frame is the SROI predefined by the FROI area. The data of the SROI area is input into a seat belt discriminator based on CNN, and the correct conclusion of "With Safety Belt" is obtained.

[0083] When the solution means according to the embodiment of the present application is not adopted, for example, when the entire target image is input into the seat belt discriminator, as shown in FIG. 8, the seat belt feature is weaker in the entire image compared to the feature of the human face, and it is difficult to be recognized by the seat belt discriminator. Therefore, in the same scene, it is difficult to judge the SROI area, and an incorrect conclusion of "Without Safety Belt" is obtained.

[0084] According to Example 1 of the present application, non-contact recognition of the seat belt wearing state of personnel based on vision is realized, the production cost of the device is saved, and at the same time, the result of judging the seat belt wearing state of personnel becomes more accurate. On the other hand, based on the size of the seat belt in the actual scene, the detection area of the seat belt is reduced by face detection, the feature of the seat belt itself is made prominent, and the recognizable degree of the seat belt wearing area is improved.

[0085] It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from the order here.

[0086] Example 1 of the present application further provides a seat belt wearing detection device. The seat belt wearing detection device according to Example 1 of the present application can be used to execute the seat belt wearing detection method provided by Example 1 of the present application. Hereinafter, the seat belt wearing detection device provided by Example 1 of the present application will be described.

[0087] Example 2 FIG. 9 is a schematic diagram of a seat belt wearing detection device according to Example 2 of the present application. As shown in FIG. 9, the device includes a first acquisition unit 10, a first identification unit 20, a second identification unit 30, and a determination unit 40.

[0088] Specifically, the first acquisition unit 10 is configured to acquire a target image collected from a transportation device.

[0089] The first identification unit 20 is configured to identify a first image region and / or a second image region in the target image. The first image region is a region where the head and torso of the target object are located in the target image, and the second image region is a region where the face of the target object is located in the target image.

[0090] The second identification unit 30 is configured to identify a third image region based on the first image region and / or the second image region. The third image region is a seat belt wearing region in the target image.

[0091] The determination unit 40 is configured to determine whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region.

[0092] The seat belt wearing detection device provided by Embodiment 2 of the present application includes a first acquisition unit 10 configured to acquire a target image collected from a transportation device, and a first identification unit 20 configured to identify a first image area and / or a second image area in the target image, where the first image area is an area where the head and torso of the target object are located in the target image, and the second image area is an area where the face of the target object is located in the target image. The first identification unit 20, and a second identification unit 30 configured to identify a third image area based on the first image area and / or the second image area, where the third image area is the seat belt wearing area in the target image. The second identification unit 30, and a determination unit 40 configured to determine whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area. By this, the problem that it is difficult to accurately determine whether a passenger is wearing a seat belt correctly in the related art is solved. By identifying the seat belt wearing area based on the area where the head and torso of the target object are located and / or the area where the face is located in the target image, it is determined whether the wearing of the seat belt meets a predetermined wearing requirement, and further, the accuracy of determining whether a passenger is wearing a seat belt correctly is improved.

[0093] Preferably, in the seat belt wearing detection device provided by Embodiment 2 of the present application, when the target image includes the upper body area of the target object, the area where the head and torso of the target object are located is the smallest area including the face, neck, shoulders and upper chest, and the area where the face of the target object is located is the smallest area including the face.

[0094] Preferably, in the seat belt wearing detection device provided by Example 2 of the present application, before acquiring a target image collected from a vehicle, the device acquires a plurality of images collected from the vehicle, marks whether the acquired images include a head and torso region, and marks the position of the head and torso region for the images including the head and torso region, and is configured to obtain a plurality of first marked images; a second acquisition unit; identifying a plurality of first marked images as first training set data, and training a neural network model with the first training set data to obtain a first detector for detecting a head and torso region; a first training unit; and / or acquiring a plurality of images collected from a vehicle, marking whether the acquired images include a face region, and marking the position of the face region for the images including the face region, and is configured to obtain a plurality of second marked images; a third acquisition unit; identifying a plurality of second marked images as second training set data, and training a neural network model with the second training set data to obtain a second detector for detecting a face region; a second training unit.

[0095] Preferably, in the seat belt wearing detection device provided by Example 2 of the present application, the first specific unit 20 includes a first specific module and / or a second specific module. When the first specific module obtains a first detector and a second detector, it is configured to detect a target image by the first detector and obtain a first image region. The second specific module is configured to detect a target image by the second detector and obtain a second image region. When the first specific module obtains the first detector, it is configured to detect a target image by the first detector and obtain a first image region. The second specific module is configured to cut out the second image region from the first image region based on the predetermined sitting posture information of the target object. When the second specific module obtains the second detector, it is configured to detect a target image by the second detector and obtain a second image region. The first specific module is configured to expand the second image region downward, leftward, and rightward by a predetermined region in the target image based on the predetermined sitting posture information of the target object to obtain the first image region.

[0096] Preferably, in the seat belt wearing detection device provided by Example 2 of the present application, when the second specific unit 30 obtains the first image region in the target image and the second image region in the target image, a third specific module configured to identify a first seat belt wearing region in the target image based on an offset distance of a covering region with respect to the first image region when the seat belt is correctly worn, and to identify a second seat belt wearing region in the target image based on an offset distance of the covering region with respect to the second image region when the seat belt is correctly worn, and a fourth specific module configured to average the first seat belt wearing region and the second seat belt wearing region at a spatial position of the target image to obtain a third image region.

[0097] Preferably, in the seat belt wearing detection device provided by Embodiment 2 of the present application, before specifying the third image area based on the first image area and / or the second image area, the device acquires a plurality of head and torso area images, marks whether the acquired images include a seat belt wearing area, and marks the position of the face area for the images including the seat belt wearing area, so as to obtain a plurality of third marked images. The fourth acquisition unit is configured to specify a plurality of third marked images as third training set data, train a neural network model with the third training set data, and obtain a third detector for detecting the seat belt wearing area. The second specifying unit is further configured to input the position of the target image in the first image area, the position of the target image in the second image area, and the first image area into the third detector to detect the third image area.

[0098] Preferably, in the seat belt wearing detection device provided by Embodiment 2 of the present application, before specifying the third image area based on the first image area and / or the second image area, the device acquires a plurality of head and torso area images, marks whether the acquired images include a seat belt wearing area, and marks the position of the seat belt wearing area for the images including the seat belt wearing area, so as to obtain a plurality of fourth marked images. The fifth acquisition unit is configured to specify a plurality of fourth marked images as fourth training set data, train a neural network model with the fourth training set data, and obtain a fourth detector for detecting the seat belt wearing area. The second specifying unit is further configured to detect the first image area by the fourth detector and obtain the third image area.

[0099] Preferably, in the seat belt wearing detection device provided by Embodiment 2 of the present application, before identifying the third image area based on the first image area and / or the second image area, the device acquires a plurality of images including the facial area, marks whether the acquired images include the seat belt wearing area, and marks the position of the seat belt wearing area for the images including the seat belt wearing area, so as to obtain a plurality of fifth marked images. The device further includes a sixth acquisition unit configured to identify a plurality of fifth marked images as fifth training set data, and train a neural network model with the fifth training set data to obtain a fifth detector for detecting the seat belt wearing area. The second identification unit is further configured to detect the second image area by the fifth detector and obtain the third image area.

[0100] Preferably, in the seat belt wearing detection device provided by Embodiment 2 of the present application, before determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area, the device acquires a plurality of seat belt wearing area images, marks whether the wearing of the seat belt in the acquired images meets the predetermined wearing requirement, so as to obtain a plurality of sixth marked images. The device further includes a seventh acquisition unit configured to identify a plurality of sixth marked images as sixth training set data, and train a neural network model with the sixth training set data to obtain a discriminator for determining whether the wearing of the seat belt in the target image meets the predetermined wearing requirement. The determination unit 40 is further configured to detect the third image area by the discriminator and obtain a conclusion on whether the wearing of the seat belt meets the predetermined wearing requirement.

[0101] Preferably, in the seat belt wearing detection device provided by Example 2 of the present application, after determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area, if the wearing of the seat belt does not meet the predetermined wearing requirement, the information issuing unit is further configured to issue prompting information for prompting the passenger to correctly wear the seat belt.

[0102] The seat belt wearing detection device includes a processor and a memory. The first acquisition unit 10, the first identification unit 20, the second identification unit 30, the judgment unit 40, etc. are all stored in the memory as program units, and the corresponding functions are realized by the processor executing the program units stored in the memory.

[0103] The processor includes a kernel, and the kernel calls the corresponding program unit from the memory. One or more kernels may be provided, and by adjusting the kernel parameters, the problem that it is difficult to accurately determine whether a passenger in related art is correctly wearing a seat belt is solved.

[0104] The memory may include forms such as volatile memory, random access memory (RAM) and / or non-volatile memory in a computer-readable medium. For example, it is a read-only memory (ROM) or flash memory (flash RAM), and the memory includes at least one storage chip.

[0105] The embodiment of the present application further provides a non-volatile storage medium, and the non-volatile storage medium includes a stored program. Here, when the program is executed, the non-volatile storage medium is arranged to control a device for executing a seat belt wearing detection method.

[0106] Embodiments of the present application include a processor and a memory, with computer-readable instructions stored in the memory, and the processor is arranged to execute the computer-readable instructions. When the computer-readable instructions are executed, an electronic device for executing a seat belt wearing detection method is further provided. The electronic device in this specification may be a server, a PC, a PAD, a mobile phone, etc.

[0107] It will be apparent to those skilled in the art that embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. And the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0108] The present application will be described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be realized by computer program instructions. By providing these computer program instructions to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, an apparatus is generated for realizing the functions specified in one or more flows of the flowchart and / or one or more blocks of the block diagram by instructions executed by the processor of the computer or other programmable data processing device.

[0109] These computer program instructions may be stored in a computer-readable memory that can guide a computer or other programmable data processing apparatus to operate in a specific manner, whereby the instructions stored in the computer-readable memory generate a manufactured article including an instruction device for realizing the functions specified in one or more flows of the flowchart and / or one or more blocks of the block diagram.

[0110] These computer program instructions may be loaded into a computer or other programmable data processing apparatus, whereby a series of operation steps are executed on the computer or other programmable apparatus to generate a process implemented by the computer, and thus the instructions executed on the computer or other programmable apparatus provide steps for realizing the functions specified in one or more processes of the flowchart and / or one or more blocks of the block diagram.

[0111] In a typical arrangement, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and a memory.

[0112] The memory may include forms such as volatile memory, random access memory (RAM) and / or non-volatile memory in a computer-readable medium, for example, read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0113] A computer-readable medium includes volatile and non-volatile, removable and non-removable media and can implement information storage by any method or technology. The information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only compact disc memory (CD-ROM), digital versatile disc (DVD) or other optical storage devices, cassette magnetic tape, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media, and can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory media such as modulated data signals and carriers.

[0114] Note that the terms "include", "including" or any variation thereof are intended to cover non-exclusive inclusion, whereby a process, method, article or apparatus that includes a series of elements includes not only those elements but also other elements not expressly listed or elements specific to those processes, methods, articles or apparatus. Without more limitations, an element limited by the phrase "comprising one..." does not exclude the presence of other like elements in the process, method, article or apparatus that includes the element.

[0115] It will be apparent to those skilled in the art that the embodiments of the present application can be provided as a method, a system, or a computer program product. Accordingly, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. And the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

[0116] The above are only embodiments of the present application and do not limit the present application. For those skilled in the art, various changes and modifications are possible to the present application. Any corrections, equivalent replacements, improvements, etc. made within the spirit and principle of the present application should all be included within the scope of the claims of the present application.

Industrial Applicability

[0117] The solution provided by the embodiments of the present invention can be applied to the technical field of image recognition. In the embodiments of the present invention, acquiring a target image collected from a transportation device, and identifying a first image region and / or a second image region in the target image, where the first image region is a region where the head and torso of the target object in the target image are located, and the second image region is a region where the face of the target object in the target image is located, and identifying a third image region based on the first image region and / or the second image region, where the third image region is a seat belt wearing region in the target image, and determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region, including, the seat belt wearing detection according to the embodiments of the present invention solves the problem that it is difficult to accurately determine whether a passenger is wearing a seat belt correctly in the related art.

Claims

1. Obtaining a target image collected from a transportation device, Identifying a first image region and a second image region in the target image, where the first image region is a region where the head and torso of the target object are located in the target image, and the second image region is a region where the face of the target object is located in the target image; and identifying a third image region based on the first image region and the second image region, where the third image region is a seat belt wearing region in the target image, Judging whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image region, A seat belt wearing detection method, Identifying the third image region based on the first image region and the second image region includes, When obtaining the first image region in the target image and obtaining the second image region in the target image, identifying a first seat belt wearing region in the target image based on the offset distance of the covering region with respect to the first image region when the seat belt is correctly worn, and identifying a second seat belt wearing region in the target image based on the offset distance of the covering region with respect to the second image region when the seat belt is correctly worn, And averaging the first seat belt wearing region and the second seat belt wearing region at the spatial position of the target image to obtain the third image region. A seat belt wearing detection method.

2. When the upper body region of the target object appears in the target image, the region where the head and torso of the target object are located is the smallest region including the face, neck, shoulders and upper chest, and the region where the face of the target object is located is the smallest region including the face. The method according to claim 1.

3. Before obtaining the target image collected from the transportation device, the method includes, Acquire a plurality of images collected from the transportation device, mark whether the acquired images include a head and torso region, and mark the position of the head and torso region for the images including the head and torso region, to obtain a plurality of first marked images, Specify the plurality of first marked images as first training set data, train a neural network model with the first training set data, and obtain a first detector configured to detect the head and torso region, and / or, Acquire a plurality of images collected from the transportation device, mark whether the acquired images include a face region, and mark the position of the face region for the images including the face region, to obtain a plurality of second marked images, Specify the plurality of second marked images as second training set data, train a neural network model with the second training set data, and obtain a second detector configured to detect the face region, and further include The method according to claim 1.

4. When obtaining the first detector and the second detector, specifying the first image region in the target image includes detecting the target image by the first detector to obtain the first image region, and specifying the second image region in the target image includes detecting the target image by the second detector to obtain the second image region, When obtaining the first detector, specifying the first image region in the target image includes detecting the target image by the first detector to obtain the first image region, and specifying the second image region in the target image includes cutting out the second image region from the first image region based on the predetermined sitting posture information of the target object, When the second detector is obtained, identifying the second image area in the target image includes detecting the target image with the second detector to obtain the second image area, and identifying the first image area in the target image includes expanding the second image area in the target image downward, leftward, and rightward by a predetermined area based on the predetermined sitting posture information of the target object to obtain the first image area. The method according to claim 3.

5. Before identifying the third image area based on the first image area and the second image area, the method includes: Acquiring a plurality of head and torso area images, marking whether the acquired images include the seat belt wearing area, and marking the position of the face area for the images including the seat belt wearing area to obtain a plurality of third marked images; Identifying the plurality of third marked images as third training set data, training a neural network model with the third training set data, and obtaining a third detector configured to detect the seat belt wearing area. When the first image area in the target image is obtained and the second image area in the target image is obtained, identifying the third image area based on the first image area and the second image area includes: Inputting the position of the first image area in the target image, the position of the second image area in the target image, and the first image area into the third detector to detect the third image area. The method according to claim 1.

6. Before identifying the third image area based on the first image area and the second image area, the method includes: Acquire a plurality of head and torso region images, mark whether the acquired images include the seat belt wearing region, and mark the position of the seat belt wearing region for the images including the seat belt wearing region, to obtain a plurality of fourth marked images, further include identifying the plurality of fourth marked images as fourth training set data, training a neural network model with the fourth training set data, and obtaining a fourth detector configured to detect the seat belt wearing region, When the first image region is detected in the target image, identifying the third image region based on the first image region and the second image region, includes detecting the first image region by the fourth detector and obtaining the third image region The method according to claim 1.

7. Before identifying the third image region based on the first image region and the second image region, the method, acquires a plurality of images including a facial region, marks whether the acquired images include the seat belt wearing region, and marks the position of the seat belt wearing region for the images including the seat belt wearing region, to obtain a plurality of fifth marked images, further includes identifying the plurality of fifth marked images as fifth training set data, training a neural network model with the fifth training set data, and obtaining a fifth detector configured to detect the seat belt wearing region, When the second image region is detected in the target image, identifying the third image region based on the first image region and the second image region, includes detecting the second image region by the fifth detector and obtaining the third image region The method according to claim 1.

8. Before determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area, the method includes: Acquiring a plurality of seat belt wearing area images, marking whether the wearing of the seat belt in the acquired images meets a predetermined wearing requirement, and obtaining a plurality of sixth marked images; Identifying the plurality of sixth marked images as sixth training set data, training a neural network model with the sixth training set data, and obtaining a discriminator configured to determine whether the wearing of the seat belt in the target image meets a predetermined wearing requirement; Based on the image information in the third image area, determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement includes: Detecting the third image area by the discriminator and obtaining a conclusion as to whether the wearing of the seat belt meets a predetermined wearing requirement. The method according to claim 1.

9. After determining whether the wearing of the seat belt in the target image meets a predetermined wearing requirement based on the image information in the third image area, the method further includes: When the wearing of the seat belt does not meet a predetermined wearing requirement, issuing prompting information configured to prompt a passenger to wear the seat belt correctly. The method according to claim 1.

10. A first acquisition unit configured to acquire a target image collected from a transportation device; A first identification unit configured to identify a first image area and a second image area in the target image, where the first image area is an area where the head and torso of a target object in the target image are located, and the second image area is an area where the face of the target object in the target image is located. A second specifying unit configured to specify a third image area based on the first image area and the second image area, wherein the third image area is a seat belt wearing area in the target image. A determination unit configured to determine whether wearing of the seat belt in the target image satisfies a predetermined wearing requirement based on image information in the third image area, and includes: A seat belt wearing detection device, The second specifying unit that specifies the third image area based on the first image area and the second image area is: When the first image area in the target image is obtained and the second image area in the target image is obtained, based on an offset distance of a covering area with respect to the first image area when the seat belt is correctly worn, a first seat belt wearing area is specified in the target image, and based on an offset distance of the covering area with respect to the second image area when the seat belt is correctly worn, a second seat belt wearing area is specified in the target image, and the first seat belt wearing area and the second seat belt wearing area are averaged in a spatial position of the target image to obtain the third image area. A seat belt wearing detection device.

11. A non-volatile storage medium including a stored program, When the program is executed, the non-volatile storage medium is arranged to control a device that executes the seat belt wearing detection method according to any one of claims 1 to 9. A non-volatile storage medium.

12. An electronic device including a processor and a memory, wherein computer-readable instructions are stored in the memory, and the processor is configured to execute the computer-readable instructions. When the computer-readable instructions are executed, the seat belt wearing detection method according to any one of claims 1 to 9 is executed. An electronic device.

Images