Method, device and computer readable storage medium for automatically identifying a car accident

Abstract

The application discloses a method, device and computer readable storage medium for automatically identifying a car accident, which comprises the following steps: determining an abnormal vehicle according to a snapshot image of a target vehicle; performing a time-lapse snapshot on a region where the target vehicle is located when the target vehicle is determined as the abnormal vehicle; and judging whether the car accident occurs or not according to the time-lapse snapshot image. The method, device and computer readable storage medium for automatically identifying the car accident can automatically and intuitively identify whether the car accident occurs or not and a vehicle damage condition, thereby laying a foundation for alarm and rescue.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic accident information, and in particular to a method and device for automatically identifying a car accident and a computer readable storage medium. BACKGROUND

[0002] Currently, whether a vehicle has been involved in a traffic accident is determined by detection results such as sensors, acceleration sensors, pressure sensors, etc. However, this method cannot determine whether a vehicle has been involved in a collision and the degree of the collision in an intuitive manner. SUMMARY

[0003] The present application provides a method and device for automatically identifying a car accident and a computer readable storage medium, which can automatically and intuitively identify whether a car accident has occurred and the damage to a vehicle, thereby laying a foundation for alarm and rescue.

[0004] The present application provides a method for automatically identifying a car accident, which comprises:

[0005] determining an abnormal vehicle according to a captured image of a target vehicle;

[0006] delayed capturing an area where the target vehicle is located when the target vehicle is determined to be an abnormal vehicle;

[0007] judging whether a car accident has occurred according to a delayed captured image.

[0008] In an exemplary embodiment, determining an abnormal vehicle according to a captured image of a target vehicle comprises:

[0009] three-dimensionally modeling the captured image of the target vehicle;

[0010] comparing the three-dimensionally modeled target vehicle with a model of a vehicle model library to obtain a deformation level of the target vehicle; and if the deformation level of the target vehicle exceeds a first preset threshold, determining the target vehicle to be an abnormal vehicle.

[0011] In an exemplary embodiment, before determining an abnormal vehicle according to a captured image of a target vehicle, the method further comprises:

[0012] obtaining a plurality of accident vehicle sample images, and calculating an overall deformation degree of an accident vehicle corresponding to each accident vehicle sample image;

[0013] dividing accident vehicle deformation levels according to overall deformation degrees of all accident vehicles;

[0014] establishing a vehicle model library according to the accident vehicle deformation levels and the corresponding accident vehicle sample images;

[0015] The method for calculating the overall deformation degree of the accident vehicle comprises,

[0016] establishing a three-dimensional model of the accident vehicle, dividing the accident vehicle into n continuous units in the three-dimensional space; wherein n is a positive integer;

[0017] counting the number of deformed units, and determining the deformation amount of each deformed unit;

[0018] calculating the overall deformation degree of the accident vehicle according to the number of deformed units and the deformation amount of the deformed units of the accident vehicle.

[0019] In an exemplary embodiment, determining an abnormal vehicle according to the image of the target vehicle captured by the camera, comprising:

[0020] three-dimensionally modeling the image of the target vehicle captured by the camera;

[0021] calculating the overall deformation degree of the target vehicle according to the three-dimensional modeling of the target vehicle; and determining the target vehicle as an abnormal vehicle if the overall deformation degree of the target vehicle exceeds a second preset threshold.

[0022] In an exemplary embodiment, determining whether a car accident has occurred according to the image captured by the time-lapse camera, comprising:

[0023] if there is no target vehicle in the image captured by the time-lapse camera and there is no scattered object on the ground in the image captured by the time-lapse camera, determining that no car accident has occurred;

[0024] if there is a target vehicle in the image captured by the time-lapse camera, determining that a car accident has occurred.

[0025] In an exemplary embodiment, determining whether a car accident has occurred according to the image captured by the time-lapse camera, comprising:

[0026] if there is a scattered object on the ground in the image captured by the time-lapse camera, determining whether the scattered object on the ground belongs to the vehicle parts of an abnormal vehicle, and determining that a car accident has occurred if the scattered object belongs to the vehicle parts of an abnormal vehicle, otherwise determining that no car accident has occurred.

[0027] In an exemplary embodiment, the determination of whether the scattered object on the ground belongs to the vehicle parts of an abnormal vehicle, comprising:

[0028] comparing the scattered object on the ground captured by the time-lapse camera with the vehicle parts model of the abnormal vehicle in the pre-established vehicle parts model library;

[0029] if the similarity between the scattered object on the ground captured by the time-lapse camera and the vehicle parts model of the abnormal vehicle in the pre-established vehicle parts model library is greater than a preset threshold, determining that the scattered object on the ground in the image captured by the time-lapse camera belongs to the vehicle parts of an abnormal vehicle.

[0030] In an example embodiment, the vehicle part model library is established in the following way:

[0031] Images of the same vehicle part from different angles and images of the same vehicle part from different brands are selected, and the selected images of the vehicle part are subjected to vehicle part type identification;

[0032] According to the vehicle part type and the corresponding vehicle part image, the vehicle part model library is established.

[0033] The application provides an automatic vehicle accident identification device, comprising:

[0034] The device comprises an abnormal vehicle determination module, a time-lapse shooting module and a vehicle accident judgment module;

[0035] The abnormal vehicle determination module is configured to determine an abnormal vehicle according to the image of the target vehicle;

[0036] The time-lapse shooting module is configured to perform time-lapse shooting on the area where the target vehicle is located when the target vehicle is determined to be an abnormal vehicle;

[0037] The vehicle accident judgment module is configured to judge whether a vehicle accident has occurred according to the time-lapse shot image.

[0038] The application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the automatic vehicle accident identification method.

[0039] Other features and advantages of the application will be described in the following description, and some will become apparent from the description, or will be understood from the practice of the application. Other advantages of the application can be achieved and obtained by the schemes described in the specification and the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0040] The accompanying drawings are used to provide an understanding of the technical solutions of the application, and constitute a part of the specification, and are used to explain the technical solutions of the application together with the embodiments of the application, and do not constitute a limitation on the technical solutions of the application.

[0041] Figure 1 A schematic diagram of the automatic vehicle accident identification method of the embodiment of the application;

[0042] Figure 2 A flowchart of the automatic vehicle accident identification method of the embodiment of the application;

[0043] Figure 3 A process model diagram example of the automatic vehicle accident identification of the embodiment of the application;

[0044] Figure 4An example of the automatic identification of the process model of the car accident of the embodiment of the present application;

[0045] Figure 5 An example of the automatic identification of the process model of the car accident of the embodiment of the present application;

[0046] Figure 6 A schematic diagram of the device for automatically identifying the car accident of the embodiment of the present application. DETAILED DESCRIPTION

[0047] Figure 1 A flow chart of the method for automatically identifying the car accident of the embodiment of the present application, as shown in the figure, the method for automatically identifying the car accident of the embodiment of the present application comprises steps S11-S13: Figure 1

[0048] S11, determining an abnormal vehicle according to the image of the target vehicle captured;

[0049] S12, performing a time-lapse capture on the area where the target vehicle is determined to be the abnormal vehicle;

[0050] S13, judging whether a car accident has occurred according to the image of the time-lapse capture.

[0051] In an exemplary embodiment, determining an abnormal vehicle according to the image of the target vehicle captured comprises:

[0052] performing three-dimensional modeling on the image of the target vehicle captured;

[0053] comparing the three-dimensional modeling of the target vehicle with the model of the vehicle model library established in advance to obtain the deformation level of the target vehicle; if the deformation level of the target vehicle exceeds a first preset threshold, the target vehicle is determined to be the abnormal vehicle.

[0054] In an exemplary embodiment, before determining an abnormal vehicle according to the image of the target vehicle captured, the method further comprises:

[0055] obtaining a plurality of accident vehicle sample images, and calculating the overall deformation degree of the accident vehicle corresponding to each accident vehicle sample image;

[0056] dividing the deformation level of the accident vehicle according to the overall deformation degree of all the accident vehicles;

[0057] establishing a vehicle model library according to the deformation level of the accident vehicle and the corresponding accident vehicle sample image;

[0058] The method for calculating the overall deformation degree of the accident vehicle comprises,

[0059] establishing a three-dimensional model of the accident vehicle, and dividing the accident vehicle into n continuous units in the three-dimensional space; wherein n is a positive integer; ​

[0060] counting the number of deformed units, and determining the deformation amount of each deformed unit;

[0061] calculating the overall deformation degree of the accident vehicle according to the number of deformed units and the deformation amount of the deformed units of the accident vehicle.

[0062] By establishing a vehicle model library, the target vehicle can be compared with the models in the vehicle model library, which not only realizes the unification of the judgment standard, but also improves the efficiency of judging whether the target vehicle is an abnormal vehicle.

[0063] In an exemplary embodiment, determining an abnormal vehicle according to a snapshot image of a target vehicle includes:

[0064] performing three-dimensional modeling on the snapshot image of the target vehicle;

[0065] calculating the overall deformation degree of the target vehicle according to the three-dimensional modeling of the target vehicle; and determining the target vehicle as an abnormal vehicle if the overall deformation degree of the target vehicle exceeds a second preset threshold.

[0066] In an exemplary embodiment, the overall deformation degree of the target vehicle can be calculated by using the above-mentioned accident vehicle overall deformation degree calculation method.

[0067] In an exemplary embodiment, judging whether a car accident has occurred according to a time-lapse snapshot image includes:

[0068] if there is no target vehicle in the time-lapse snapshot image and there is no scattered object on the ground in the time-lapse snapshot image, determining that no car accident has occurred;

[0069] if there is a target vehicle in the time-lapse snapshot image, determining that a car accident has occurred.

[0070] In an exemplary embodiment, the time-lapse time of the time-lapse snapshot is set according to actual needs, such as considering the situation of a vehicle waiting for a red light or a yellow light.

[0071] In an exemplary embodiment, judging whether a car accident has occurred according to a time-lapse snapshot image includes:

[0072] if there is a scattered object on the ground in the time-lapse snapshot image, determining whether the scattered object on the ground belongs to a vehicle part of an abnormal vehicle;

[0073] when it is determined that the scattered object belongs to a part of an abnormal vehicle, determining that a car accident has occurred;

[0074] when it is determined that the scattered object does not belong to a part of an abnormal vehicle, determining that no car accident has occurred.

[0075] In an example embodiment, the judging whether the scatterings appearing on the ground belong to vehicle parts of the abnormal vehicle comprises:

[0076] Comparing the scatterings appearing on the ground in the time-lapse photograph with the vehicle part model of the abnormal vehicle in the vehicle part model library;

[0077] When the similarity between the scatterings appearing on the ground in the time-lapse photograph and the vehicle part model of the abnormal vehicle in the vehicle part model library is greater than the preset threshold, it is determined that the scatterings appearing on the ground in the time-lapse photograph belong to the vehicle parts of the abnormal vehicle.

[0078] When the similarity between the scatterings appearing on the ground in the time-lapse photograph and the vehicle part model of the abnormal vehicle in the vehicle part model library is less than or equal to the preset threshold, it is determined that the scatterings appearing on the ground in the time-lapse photograph do not belong to the vehicle parts of the abnormal vehicle.

[0079] In an example embodiment, the vehicle part model library is established by the following method:

[0080] Selecting images of the same vehicle part at different angles and images of the same vehicle part of different brands, and performing vehicle part type identification on the selected vehicle part images;

[0081] According to the vehicle part type and the corresponding vehicle part image, the vehicle part model library is established.

[0082] In an example embodiment, the vehicle part type identification on the selected vehicle part images comprises:

[0083] For each selected vehicle part, preprocessing and labeling are performed, and based on the TensorFlow framework, machine learning is performed using a data flow programming system on the preprocessed and labeled vehicle part image to obtain the type to which the vehicle part belongs.

[0084] The automatic vehicle accident identification method of the embodiment of the application determines the abnormal vehicle through the image of the target vehicle, performs time-lapse photographing on the area where the target vehicle is located when the target vehicle is determined to be an abnormal vehicle, and judges whether a vehicle accident has occurred according to the time-lapse photographing image, so that whether a vehicle accident has occurred and the vehicle damage condition can be automatically and intuitively identified, thereby laying a foundation for alarm and rescue.

[0085] Figure 2 As shown in the flowchart of the automatic vehicle accident identification method of the embodiment of the application, Figure 2 the flowchart comprises steps S21-S25:

[0086] S21, capturing the vehicle by the camera of the monitoring area;

[0087] S22, comparing the captured vehicle with the vehicle model library, if the captured vehicle is not an abnormal vehicle, determining that no car accident occurs, if the captured vehicle is an abnormal vehicle, executing step S23;

[0088] S23, capturing the abnormal vehicle by the camera of the monitoring area; if the abnormal vehicle is captured, determining that the car accident occurs; if the scattered object on the ground is captured, executing step S24; if no abnormal vehicle is captured and no scattered object on the ground is captured, determining that no car accident occurs;

[0089] S24, comparing the scattered object with the vehicle parts in the vehicle parts library;

[0090] S25, if the similarity exceeds the preset threshold, determining that the scattered object is the vehicle part of the abnormal vehicle, and determining that the car accident occurs; if it is determined that the scattered object is not the vehicle part, determining that no car accident occurs.

[0091] In step S21, the intersection can be captured by the camera, or other monitoring areas can be captured.

[0092] In step S22, the vehicle model library can collect case pictures of automobile collision accidents for learning, and the local deformation area of the accident vehicle is divided into grids by using the finite element method, hexahedral mesh generation algorithm and ten-node curved quadrilateral. For each grid element, a collision force calculation model and a collision instantaneous speed output model are established based on vehicle dynamics and elastic mechanics theory.

[0093] The specific steps are as follows:

[0094] S221, establishing a three-dimensional model of the accident vehicle, and dividing the accident vehicle into n continuous units in the three-dimensional space;

[0095] S222, measuring the number of units deformed in the deformation area of the accident vehicle and the corresponding deformation amount of each deformed unit, and marking them in the three-dimensional model;

[0096] S223, taking the deformation amount of each unit as a training sample, taking the deformation coefficient as an output result, calculating the deformation coefficient of each deformed unit by using the convolution neural network algorithm, and then weighting and averaging each deformed unit according to the number of deformed units and the deformation coefficient of each unit to obtain the overall deformation ratio of each vehicle;

[0097] S224, select a plurality of accident vehicles of the same model as the accident vehicle, repeat steps S221-S223, and obtain the deformation coefficient of each unit and the deformation ratio of the whole vehicle of each accident vehicle through training;

[0098] S225, according to the deformation ratio of each vehicle, set M levels and the corresponding ratio of each level, M is a positive integer.

[0099] S226, three-dimensional modeling of the snapshot camera vehicle snapshot picture and comparison with the accident three-dimensional vehicle model library, according to the comparison result, the deformation ratio of the whole vehicle is calculated.

[0100] In step S221, n is a positive integer, which can be set according to actual needs.

[0101] In step S225, for example, M can be set to 6, and the six levels can be 20%, 35%, 50%, 65%, 80% and 95%.

[0102] By establishing the vehicle model library, the target vehicle can be compared with the model in the vehicle model library, which not only realizes the unification of the judgment standard, but also improves the efficiency of judging whether the target vehicle is an abnormal vehicle.

[0103] In other embodiments, when judging whether the target vehicle is an abnormal vehicle, a vehicle model library can also be established, and the overall deformation degree of the target vehicle is calculated according to the three-dimensional modeling of the target vehicle; if the overall deformation degree of the target vehicle exceeds the second preset threshold, the target vehicle is determined as an abnormal vehicle.

[0104] In step S23, since the vehicle stops in the region after the collision (as shown in Figure 3 ), the accident is judged, and then the alarm is processed in time. The delay time of the delay snapshot can be set according to actual needs, such as considering the situation of waiting for red light or yellow light when monitoring the intersection.

[0105] In step S23, the damaged vehicle has already driven out of the region, and the vehicle is not captured after the delay snapshot (as shown in Figure 4 ), it is directly determined that no accident has occurred.

[0106] In step S24, although the vehicle has driven out of the region after the collision, part of the automobile parts remains in the region (as shown in Figure 5 ), it is determined that an accident has occurred, and then the alarm is processed in time.

[0107] In step S24, the vehicle parts are identified according to the following algorithm:

[0108] The algorithm is based on the TensorFlow framework and uses a dataflow programming system for machine learning.

[0109] First, the image processing method is selected, the image is augmented, the color of the image is randomly changed through geometric transformation, the size of the components of the three channels of RGB is scaled according to the proportion, and the size, contrast and brightness are transformed, some light is added to the picture, such as sharpening, convex point, adaptive histogram equalization, then the picture is horizontally and vertically flipped, scaled, rotated, part of the picture is randomly cropped from the original image, noise and image blur processing is added.

[0110] Then the parts of the car are classified, such as front bumper, rear bumper, door handle, front and rear door, etc., and then the pictures of each category are labeled: the pictures of different angles and different brands of the same category of parts are identified by the neural network and the model is trained. Through training, the classification probability of each part is calculated, and the corresponding classification probability is calculated, so as to select which category.

[0111] Finally, the scattered objects on the ground and the car part model are compared, and if the similarity is higher than 90%, it is considered that the part is scattered.

[0112] The automatic identification of the method for identifying the car accident of the embodiment of the application can automatically and intuitively identify whether a car accident has occurred and the damage to the vehicle, thereby laying a foundation for alarm and rescue.

[0113] Figure 6 The schematic diagram of the device for automatically identifying the car accident of the embodiment of the application is shown in FIG. 1. Figure 6 The device for automatically identifying the car accident of the embodiment of the application comprises an abnormal vehicle determination module, a delay snapshot module and a car accident judgment module.

[0114] The abnormal vehicle determination module is configured to determine an abnormal vehicle according to a snapshot image of a target vehicle.

[0115] The delay snapshot module is configured to delay snapshot an area where the target vehicle is located when the target vehicle is determined as an abnormal vehicle.

[0116] The car accident judgment module is configured to judge whether a car accident has occurred according to the delay snapshot image.

[0117] In an exemplary embodiment, the abnormal vehicle is determined according to the snapshot image of the target vehicle, comprising:

[0118] Three-dimensional modeling is performed on the snapshot image of the target vehicle.

[0119] The three-dimensional modeling of the target vehicle is compared with the models in the pre-established vehicle model library to obtain a deformation level of the target vehicle; if the deformation level of the target vehicle exceeds a first preset threshold, the target vehicle is determined as an abnormal vehicle.

[0120] In an exemplary embodiment, before determining the abnormal vehicle according to the image of the target vehicle captured by the snapshot, the method further comprises:

[0121] A plurality of accident vehicle sample images are obtained, and an overall deformation degree of each accident vehicle corresponding to the accident vehicle sample image is calculated;

[0122] According to the overall deformation degrees of all the accident vehicles, the accident vehicle deformation levels are classified;

[0123] According to the accident vehicle deformation levels and the corresponding accident vehicle sample images, a vehicle model library is established;

[0124] The method for calculating the overall deformation degree of the accident vehicle comprises:

[0125] A three-dimensional model of the accident vehicle is established, and the accident vehicle is divided into n continuous units in a three-dimensional space; wherein n is a positive integer;

[0126] The number of the units that have been deformed is counted, and the deformation amount of each unit that has been deformed is determined;

[0127] The overall deformation degree of the accident vehicle is calculated according to the number of the units that have been deformed and the deformation amount of the units that have been deformed.

[0128] In an exemplary embodiment, determining the abnormal vehicle according to the image of the target vehicle captured by the snapshot comprises:

[0129] The image of the target vehicle captured by the snapshot is three-dimensionally modeled;

[0130] The overall deformation degree of the target vehicle is calculated according to the three-dimensional modeling of the target vehicle; if the overall deformation degree of the target vehicle exceeds a second preset threshold, the target vehicle is determined as an abnormal vehicle.

[0131] In an exemplary embodiment, the overall deformation degree of the target vehicle can be calculated by using the above-mentioned method for calculating the overall deformation degree of the accident vehicle.

[0132] In an exemplary embodiment, determining whether a car accident has occurred according to the image captured by the time-lapse snapshot comprises:

[0133] If there is no target vehicle in the image captured by the time-lapse snapshot and there is no scattered object on the ground in the image captured by the time-lapse snapshot, it is determined that no car accident has occurred;

[0134] If the target vehicle exists in the image captured by the time-lapse photograph, it is determined that a car accident has occurred.

[0135] In an exemplary embodiment, the time-lapse time of the time-lapse photograph is set according to actual needs, such as considering the situation that a vehicle waits for a red light or a yellow light.

[0136] In an exemplary embodiment, determining whether a car accident has occurred according to the image captured by the time-lapse photograph comprises:

[0137] If the scattered object on the ground exists in the image captured by the time-lapse photograph, it is determined whether the scattered object on the ground belongs to the vehicle part of the abnormal vehicle.

[0138] When it is determined that the scattered object belongs to the vehicle part of the abnormal vehicle, it is determined that a car accident has occurred.

[0139] When it is determined that the scattered object does not belong to the vehicle part of the abnormal vehicle, it is determined that a car accident has not occurred.

[0140] In an exemplary embodiment, the determination of whether the scattered object on the ground belongs to the vehicle part of the abnormal vehicle comprises:

[0141] Comparing the scattered object on the ground captured by the time-lapse photograph with the vehicle part model of the abnormal vehicle in the pre-established vehicle part model library;

[0142] When the similarity between the scattered object on the ground captured by the time-lapse photograph and the vehicle part model of the abnormal vehicle in the pre-established vehicle part model library is greater than a preset threshold, it is determined that the scattered object on the ground in the image captured by the time-lapse photograph belongs to the vehicle part of the abnormal vehicle.

[0143] When the similarity between the scattered object on the ground captured by the time-lapse photograph and the vehicle part model of the abnormal vehicle in the pre-established vehicle part model library is less than or equal to a preset threshold, it is determined that the scattered object on the ground in the image captured by the time-lapse photograph does not belong to the vehicle part of the abnormal vehicle.

[0144] In an exemplary embodiment, the vehicle part model library is established by the following method:

[0145] Selecting images of the same vehicle part at different angles and images of the same vehicle part of different brands, and performing vehicle part type identification on the selected vehicle part images respectively;

[0146] According to the vehicle part type and the corresponding vehicle part image, a vehicle part model library is established.

[0147] In an exemplary embodiment, the vehicle part type identification on the selected vehicle part images respectively comprises:

[0148] For each selected vehicle part, pre-processing and labeling are performed; for the pre-processed and labeled vehicle part image, machine learning is performed based on a TensorFlow framework using a data flow programming system; and a type to which the vehicle part belongs is obtained.

[0149] The device for automatically identifying a car accident according to the embodiments of the present application can automatically and intuitively identify whether a car accident has occurred and a vehicle damage condition, thereby laying a foundation for alarm and rescue.

[0150] The present application also discloses a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method for automatically identifying a car accident.

[0151] The present application describes a plurality of embodiments, but the description is exemplary rather than limiting, and it is obvious to those skilled in the art that there can be more embodiments and implementation schemes within the scope of the embodiments described in the present application. Although many possible combinations of features are shown in the drawings and discussed in the detailed description, many other combinations of the disclosed features are also possible. Unless specifically limited, any feature or element of any embodiment can be used with any other feature or element of any other embodiment, or can replace any other feature or element of any other embodiment.

[0152] The present application includes and contemplates combinations of features and elements known to those skilled in the art. The embodiments, features and elements disclosed in the present application can also be combined with any conventional features or elements to form a unique inventive scheme defined by the claims. Any feature or element of any embodiment can also be combined with features or elements from other inventive schemes to form another unique inventive scheme defined by the claims. Therefore, it should be understood that any feature shown and / or discussed in the present application can be implemented alone or in any appropriate combination. Therefore, the embodiments are not limited other than according to the limitations made according to the appended claims and their equivalent replacements. In addition, various modifications and changes can be made within the scope of protection of the appended claims.

[0153] Furthermore, in describing representative embodiments, the specification can have presented the method and / or process as a particular sequence of steps. However, to the extent that the method or process depends on the performance of certain steps, the method or process should not be limited to the order of steps presented, as the steps presented in the specification are not exhaustive of all possible sequential embodiments. Other steps can be provided without departing from the spirit or scope of the present application. Accordingly, where the method and / or process steps have not been expressly concluded above, the method and / or process steps shall not be deemed limited to the particular order presented in the specification. Further, the claims should not be limited to the steps of the method and / or process in the order presented in the specification, as the skilled person can readily appreciate that the order of steps can be varied and still remain within the spirit and scope of the present application.

[0154] Those of ordinary skill in the art will appreciate that all or certain steps in the methods disclosed above, functional modules / units in the systems and apparatuses, can be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division of the functional modules / units referred to in the above description does not necessarily correspond to the division of physical components; for example, one physical component can have multiple functions, or one function or step can be performed by several physical components in cooperation. Certain components or all components can be implemented as software executed by a processor, such as a digital signal processor or a microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit. Such software can be distributed on computer readable media, which can include computer storage media (or non-transitory media) and communication media (or transitory media). As is well known to those of ordinary skill in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store the desired information and which can be accessed by a computer. Further, it should be appreciated by those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes any information delivery media.

Claims

1. A method for automatically identifying traffic accidents, characterized in that: Identify abnormal vehicles based on captured images of the target vehicle; The area where the target vehicle is identified as an abnormal vehicle is captured with a time delay. Determining whether a car accident has occurred based on time-lapse images includes: If the delayed capture image does not contain the target vehicle and there are no scattered objects on the ground in the delayed capture image, it is determined that no car accident has occurred. If the target vehicle is present in the delayed image, it is determined that a car accident has occurred; If scattered objects appear on the ground in the time-lapse captured image, it is determined whether the scattered objects belong to the parts of the abnormal vehicle. If the scattered objects belong to the parts of the abnormal vehicle, it is determined that a car accident has occurred; otherwise, it is determined that no car accident has occurred.

2. The method as described in claim 1, characterized in that: Identify abnormal vehicles based on captured images of the target vehicle, including: Perform 3D modeling on the captured images of the target vehicle; The deformation level of the target vehicle is obtained by comparing the 3D model of the target vehicle with the model in the pre-established vehicle model library; if the deformation level of the target vehicle exceeds the first preset threshold, the target vehicle is identified as an abnormal vehicle.

3. The method as described in claim 2, characterized in that, Before determining the abnormal vehicle based on the captured image of the target vehicle, the process also includes: Acquire multiple accident vehicle sample images and calculate the overall deformation degree of the accident vehicle corresponding to each accident vehicle sample image; Based on the overall deformation of all accident vehicles, the deformation levels of the accident vehicles are classified. A vehicle model library is established based on the deformation levels of the accident vehicles and their corresponding sample images. The method for calculating the overall deformation of the accident vehicle includes, A three-dimensional model of the accident vehicle is established, and the accident vehicle is divided into n consecutive units in three-dimensional space; where n is a positive integer. Count the number of deformed elements and determine the amount of deformation for each deformed element; The overall deformation degree of the vehicle is calculated based on the number of deformed units and the amount of deformation of those units.

4. The method as described in claim 1, characterized in that: Identify abnormal vehicles based on captured images of the target vehicle, including: Perform 3D modeling on the captured images of the target vehicle; The overall deformation of the target vehicle is calculated based on the 3D model of the target vehicle; if the overall deformation of the target vehicle exceeds a second preset threshold, the target vehicle is identified as an abnormal vehicle.

5. The method as described in claim 1, characterized in that: The determination of whether scattered objects on the ground belong to vehicle parts of an abnormal vehicle includes: The system compares the scattered objects on the ground captured by time delay with the vehicle component models of abnormal vehicles in a pre-established vehicle component model library. If the similarity between the scattered objects on the ground captured by the time delay and the vehicle component models of abnormal vehicles in the pre-established vehicle component model library is greater than a preset threshold, then the scattered objects appearing on the ground in the time delay captured image are determined to be vehicle components of abnormal vehicles.

6. The method as described in claim 1, characterized in that: Establish a vehicle parts model library using the following method: Images of the same vehicle component from different angles, as well as images of the same vehicle component from different brands, are selected. The selected vehicle component images are then used to identify the vehicle component type. A vehicle component model library is established based on the types of vehicle components and their corresponding images.

7. A device for automatically recognizing traffic accidents, characterized in that: The device includes an abnormal vehicle identification module, a time-delayed capture module, and a traffic accident judgment module; The abnormal vehicle determination module is configured to determine abnormal vehicles based on the captured image of the target vehicle. The delayed capture module is configured to perform delayed capture of the area where the target vehicle is located when it is determined to be an abnormal vehicle. The traffic accident detection module is configured to determine whether a traffic accident has occurred based on the delayed captured image.

8. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for automatically identifying traffic accidents as described in any one of claims 1-6.

Images