Traffic violation identification method, computer device and computer readable storage medium

Abstract

The embodiments of the present application disclose a method for identifying traffic violations, a computer device, and a computer-readable storage medium. The method for identifying traffic violations includes the following steps: acquiring an on-site image; encrypting the on-site image to obtain an image to be identified; building a traffic violation identification model, and inputting the image to be identified into the traffic violation identification model to obtain and output a traffic violation result. Therefore, the present application can obtain a traffic violation result by inputting the on-site image acquired on-site into the identification model, so that a highly accurate traffic violation result can be obtained without the need for monitoring personnel to manually observe the surveillance video in real time, so that business personnel can make decisions and operate based on the traffic violation result identified; and encryption is performed during the process of transmitting the on-site image to the traffic violation identification model to protect the data security of the on-site image and avoid data leakage.

Description

Technical Field

[0001] The present application belongs to the field of behavior recognition technology, and in particular relates to a method for identifying illegal behavior, a computer device, and a computer-readable storage medium. Background Art

[0002] To monitor and control illegal actions and behaviors at work sites, many sites (such as production lines) are equipped with cameras. Humans monitor the footage in real time and address any violations. However, this approach is inefficient and relies heavily on human resources, leading to high error rates. Furthermore, the process of sending the collected images to servers often creates the risk of data leakage due to the large amount of data involved.

[0003] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0004] Based on this, it is necessary to address the above problems and propose a method for identifying illegal behavior, a computer device and a computer-readable storage medium.

[0005] The present application solves the technical problem by adopting the following technical solutions:

[0006] The present application provides a method for identifying traffic violation behaviors, comprising the following steps: acquiring a scene image; encrypting the scene image to obtain an image to be identified; constructing a traffic violation identification model, inputting the image to be identified into the traffic violation identification model to obtain and output a traffic violation result.

[0007] In an optional embodiment of the present application, before encrypting the scene image to obtain the image to be identified, the method also includes: when acquiring the scene image, obtaining audio information collected at the same time as the image to be identified; inputting the scene image into a sound prediction model to obtain predicted audio information; judging whether the scene image is a real image based on the audio information and the predicted audio information; and encrypting the scene image when the scene image is a real image.

[0008] In an optional embodiment of the present application, before encrypting the scene image to obtain the image to be identified, the method also includes: obtaining the identity information of the operator; obtaining a public key and a private key matching the identity information based on the identity information; encrypting the scene image to obtain the image to be identified, including: extracting facial features from the scene image, and determining the identity information corresponding to the scene image based on the facial features; matching the corresponding private key based on the identity information; signing the scene image based on the private key to obtain an encrypted image, and packaging the encrypted image and the scene image into the image to be identified.

[0009] In an optional embodiment of the present application, inputting the image to be identified into the violation recognition model includes: decrypting the image to be identified according to the public key to obtain a decrypted image; when the decrypted image is consistent with the scene image, inputting the scene image into the violation recognition model.

[0010] In an optional embodiment of the present application, the image to be identified is input into a traffic violation identification model to obtain and output a traffic violation result, including: determining the identification object in the image to be identified, and identifying the key nodes of the identification object; determining the behavior information of the identification object based on the position relationship and position change of multiple key nodes; obtaining a preset traffic violation action identification mapping table, querying the traffic violation action identification mapping table based on the behavior information to determine whether the behavior information belongs to a traffic violation, and the traffic violation action identification mapping table is used to store the mapping relationship between behavior information and traffic violation behavior; when the behavior information belongs to a traffic violation, outputting the corresponding traffic violation result, and the traffic violation result includes a violation probability and a violation type attribute.

[0011] In an optional embodiment of the present application, the identification object in the image to be identified is determined, and the key nodes of the identification object are identified, including: when the key nodes are blocked, converting the image to be identified into a panoramic identification image; identifying the key nodes of the identification object based on the panoramic identification image, the key nodes including display key nodes and occlusion key nodes.

[0012] In an optional embodiment of the present application, a violation recognition model is constructed, including: obtaining a first training sample collected by a first terminal and a second training sample collected by a second terminal, the first terminal being a collection terminal for performing violation detection, and the second terminal being the remaining collection terminals within the same area as the first terminal; constructing an initial violation recognition model with the first training sample as an input item, and a reference model with the first training sample and the second training sample as input items, the violation recognition model and the reference model having the same output layer; using the first training sample to train the initial violation recognition model to obtain a first violation result, and using the first training sample and the second training sample to train the reference model to obtain a second violation result; obtaining a target deviation based on the first violation result and the second violation result, and iteratively adjusting the parameters of the initial violation recognition model based on the target deviation until the target deviation is less than a preset deviation threshold, thereby obtaining a violation recognition model.

[0013] In an optional embodiment of the present application, obtaining and outputting the violation result includes: generating warning information according to the violation result and outputting it to a corresponding warning terminal.

[0014] The present application also provides a computer device, comprising a processor and a memory: the processor is configured to execute a computer program stored in the memory to implement the aforementioned method.

[0015] The present application also provides a computer-readable storage medium storing a computer program, which implements the aforementioned method when the computer program is executed by a processor.

[0016] The embodiments of the present application have the following beneficial effects:

[0017] This application can obtain violation results by inputting on-site images into the recognition model, so that without the need for monitoring personnel to manually observe the surveillance video in real time, violation results with higher accuracy can be obtained, so that business personnel can make decisions and operations based on the violation results identified; and encryption is performed during the process of transmitting the on-site images to the violation recognition model to protect the data security of the on-site images and avoid data leakage.

[0018] The above description is only an overview of the technical solution of this application. In order to more clearly understand the technical means of this application, which can be implemented in accordance with the contents of the description, and to make the above and other purposes, features and advantages of this application more obvious and easy to understand, the following preferred embodiments are specifically described in detail with reference to the accompanying drawings. It should be understood that the above general description and the detailed description below are only exemplary and explanatory and do not limit this application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] in:

[0021] Figure 1 A flowchart of a traffic violation identification method provided in Example 1;

[0022] Figure 2 A structural block diagram of a computer device provided in Example 2;

[0023] Figure 3 A schematic diagram of the computer device module structure in one embodiment is shown. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0025] Example 1

[0026] Figure 1 This is a flow chart of a method for identifying traffic violations applied to a terminal provided in Example 1. To clearly describe the method for identifying traffic violations applied to a terminal provided in this embodiment, please refer to Figure 1 .

[0027] Step S110: Acquire on-site images.

[0028] Step S120: Encrypt the scene image to obtain the image to be identified.

[0029] In one embodiment, the execution terminal for steps S110 and S120 and the execution terminal for step S130 may not be the same device. That is, in steps S110 and S120, the on-site images are captured and processed, and then sent to another device for traffic violation identification. Specifically, the execution terminal for steps S110 and S120 can be a capture device, such as a camera device with image capture and simple data processing and transmission functions, or simply a data processing device, with other capture devices providing the image data to be identified. The specific form is not limited. In the case of a capture device, it can itself capture on-site images. In another embodiment, it can also be a data processing device, such as a unified management device for all capture devices, which is used to collect on-site images captured by all capture devices. Furthermore, the actual application scenarios of this embodiment may include, but are not limited to, monitoring pedestrians or drivers for traffic violations in public transportation, and monitoring employees for traffic violations during work at workplaces or production sites. Therefore, the acquired on-site image is the image captured by the monitoring device terminal in the implementation scenario exemplified in this embodiment. Since the purpose is to detect the violation of personnel, the corresponding person must be present in the image to be identified.

[0030] In one embodiment, before step S120: encrypting the scene image to obtain the image to be identified, the method further includes: when acquiring the scene image, acquiring audio information collected at the same time as the image to be identified; inputting the scene image into a sound prediction model to obtain predicted audio information; judging whether the scene image is a real image based on the audio information and the predicted audio information; and encrypting the scene image when the scene image is a real image.

[0031] In one embodiment, in the aforementioned violation identification method, identification is based on on-site images, meaning only image information. However, in real life, especially in production scenarios, manufacturers often maintain confidentiality and desensitize their production lines, so the credibility of the acquired images to be identified may not be high. Therefore, audio information captured simultaneously with the on-site images can be acquired. For example, this can involve converting a static image to be identified into a dynamic video, or multiple static images to be identified and audio information captured simultaneously. The image to be identified is input into a sound prediction model to obtain predicted audio information. The sound prediction model can be constructed and trained according to a preset method. For example, corresponding sound labels can be assigned to the training images, thereby training the sound prediction model so that it can obtain predicted audio information based on the image information. The authenticity of the image to be identified is determined based on the audio information and the predicted audio information. For example, a similarity can be calculated between the simultaneously acquired audio information and the predicted audio information. If the similarity exceeds a preset threshold, the acquired image to be identified is deemed authentic, and subsequent steps are executed.

[0032] In one embodiment, during the process of verifying whether a live image is authentic, timestamp information can be extracted from the image to be recorded. This timestamp information is hashed and added to the content of a leaf node. The sequentiality of the timestamp hash results in the leaf node is verified. If the sequentiality verification passes, the live image is determined to be captured in real time and is, therefore, authentic. Furthermore, after extracting facial features, the extracted features can be compared with the operator's identity information to verify whether the current operator in the production plan corresponds to the identified identity. Only after all multiple verifications have passed can the illegal action be identified, thus preventing impersonation of personnel and videos. This ensures the authenticity of the acquired image to be identified and improves the reliability and accuracy of the recognition results.

[0033] In one embodiment, before step S120: encrypting the scene image to obtain the image to be identified, the method further includes: obtaining identity information of the operator; obtaining a public key and a private key matching the identity information based on the identity information; encrypting the scene image to obtain the image to be identified, including: extracting facial features from the scene image, determining the identity information corresponding to the scene image based on the facial features; matching the corresponding private key based on the identity information; signing the scene image based on the private key to obtain an encrypted image, and packaging the encrypted image and the scene image into the image to be identified.

[0034] In one embodiment, it is understood that during the process of implementing traffic violation identification, a large amount of data will be transmitted, posing the risk of data leakage and directly impacting data security. Therefore, the scene image can be encrypted to obtain the image to be identified, thereby protecting data security. Encryption requires a key, and it is understood that the present application is aimed at identifying traffic violations by personnel. Identifying traffic violations will inevitably also identify the violators. Therefore, the identity information of the operator can be obtained in advance, including the operator's facial features. A digital digest is generated based on the facial features for registration, thereby obtaining a public key for decryption and a private key for encryption, wherein the private key is bound to the identity information corresponding to the facial features. After obtaining the scene image, the facial features of the traffic violation personnel in the scene image are extracted, and the identity information of the traffic violation personnel is determined based on these facial features. The identity information of the traffic violation personnel is bound to the private key used for encryption. After matching and obtaining the private key corresponding to the identity information, the scene image can be signed to obtain an encrypted image, and the encrypted image and the scene image can be packaged together to form the image to be identified. Even if a data leak occurs, the attacker cannot distinguish between the encrypted image and the on-site image because they cannot decrypt it, thus protecting data security and improving data safety.

[0035] In one embodiment, after extracting the facial features of the traffic violators in the scene image, the identities of multiple traffic violators can be desensitized. The scene image is signed according to the private key to obtain an encrypted image, including: signing the multiple scene images after desensitization; the scene image without the signature is the desensitized image. In the embodiment of this specification, an identity recognition model and a portrait desensitization model can be constructed; the identity recognition model can be trained using sample user images; the portrait desensitized image can be used to identify the portrait features in the sample user image and adjust the portrait features, and the identity recognition model can be used to predict the identity of the sample user image after the portrait features are adjusted. If the identity of the sample user is predicted, the portrait desensitization model is iteratively adjusted until the sample user image after the portrait features are adjusted by the portrait desensitization model is recognized by the identity recognition model with an accuracy rate lower than a threshold. Among them, adjusting the portrait features can be adjusting the grayscale and curvature of the portrait features in the sample user image, which is not specifically limited here. Thereby, the privacy of the person in the image to be identified is protected and privacy leakage is prevented.

[0036] Step S130: constructing a traffic violation recognition model, inputting the image to be recognized into the traffic violation recognition model to obtain and output the traffic violation result.

[0037] In one embodiment, step S130: inputting the image to be identified into the violation recognition model to obtain and output the violation result, including: determining the identification object in the image to be identified, identifying the key nodes of the identification object; determining the behavior information of the identification object based on the position relationship and position change of multiple key nodes; obtaining a preset violation action recognition mapping table, querying the violation action recognition mapping table based on the behavior information to determine whether the behavior information belongs to a violation behavior, and the violation action recognition mapping table is used to store the mapping relationship between behavior information and violation behavior; when the behavior information belongs to a violation behavior, outputting the corresponding violation result, and the violation result includes a violation probability and a violation type attribute.

[0038] In one embodiment, determining the identification object in the image to be identified and identifying the key nodes of the identification object include: when the key nodes are blocked, converting the image to be identified into a panoramic identification image; identifying the key nodes of the identification object based on the panoramic identification image, the key nodes including display key nodes and blocked key nodes.

[0039] In one embodiment, inputting the image to be identified into the violation recognition model to obtain the violation result can include processing the image to be identified to convert it into the physical behavior of the violator, identifying key human nodes in the physical behavior, determining the human body movements based on multiple key positional relationships, and reconstructing the behavioral information of the identified object at the time of capture based on the body movements. The behavioral information can be production line operations, such as a single processing action or a combination of processing actions. A preset violation action recognition mapping table is used to determine whether the behavioral information is a violation action. The violation action recognition mapping table is used to store the mapping relationship between behavioral information and violation actions. The final violation result output can include whether the violation is identified, as well as the violation probability, violation type, and other attributes. Furthermore, in the embodiment of identifying joint nodes, key nodes may be obscured by other objects during the operator's movement, resulting in missing information, which can affect node recognition and ultimately the violation result. Therefore, the image to be identified is not limited to images captured from a single perspective and can be a panoramic image including the identified object. In other words, images captured from other perspectives and other acquisition devices can also be used as the image to be identified. This means acquiring panoramic images from more perspectives, including the identified object, to capture more complete key nodes and accurately restore the operator's behavior at the time. These key nodes include display key nodes and occlusion key nodes. This means that after in-depth training, the violation recognition model can predict occlusion key nodes, including their location and movement. This ensures that even with reliable information, the operator's behavior can be fully restored, improving recognition accuracy and adapting to more complex implementation scenarios.

[0040] In one embodiment, before identifying the image to be identified, its authenticity can be identified. For example, the image hash result after hashing the image to be identified is obtained. The leaf node of the Merkle hash tree is created by combining the random number, the corresponding block node address and the image hash result, and the path information is generated according to the position of the leaf node in the Merkle hash tree as the index information of the image hash result. When verifying the authenticity of the image to be verified, the path information is used to query in the Merkle hash tree to obtain the leaf node, and the image to be verified is hashed and compared with the leaf node information. If they are consistent, the verification is passed, proving that the image to be identified corresponding to the on-site image is a real image, so as to execute the subsequent steps. Avoid the behavior of human impersonation and video impersonation, thereby ensuring the authenticity of the acquired image to be identified and improving the reliability and accuracy of the recognition results.

[0041] In one embodiment, step S130: inputting the image to be identified into the violation recognition model includes: decrypting the image to be identified according to the public key to obtain a decrypted image; when the decrypted image is consistent with the scene image, inputting the scene image into the violation recognition model.

[0042] In one embodiment, as previously described, since valid data cannot be determined before decryption, the data needs to be decrypted. The decryption process specifically involves obtaining a registered public key, using the public key to decrypt the signature signatures signed with the private key, and obtaining decrypted signatures. The decrypted signatures are then compared with the intermediate signatures. If the decrypted signatures match the intermediate signatures, indicating that the corresponding intermediate signatures were correctly obtained, the intended steps can be executed, namely, inputting the intermediate signatures into the backend processing layer to obtain the violation result.

[0043] In one embodiment, step S130: constructing a violation recognition model includes: obtaining a first training sample collected by a first terminal and a second training sample collected by a second terminal, where the first terminal is a collection terminal for violation detection, and the second terminal is the remaining collection terminals within the same area as the first terminal; constructing an initial violation recognition model with the first training sample as an input item, and a reference model with the first training sample and the second training sample as input items, the violation recognition model and the reference model having the same output layer; using the first training sample to train the initial violation recognition model to obtain a first violation result, and using the first training sample and the second training sample to train the reference model to obtain a second violation result; obtaining a target deviation based on the first violation result and the second violation result, and iteratively adjusting the parameters of the initial violation recognition model based on the target deviation until the target deviation is less than a preset deviation threshold, thereby obtaining a violation recognition model.

[0044] In one embodiment, as described above for the implementation method of identifying joint nodes, key nodes may be blocked by other objects during the movement of the operator, resulting in information loss, thereby affecting the identification of the nodes and ultimately affecting the identification of the violation results. Therefore, when constructing a violation recognition model, you can consider how to predict the position of the blocked key nodes. Therefore, you can obtain a first training sample collected by the first terminal and a second training sample collected by the second terminal. The first terminal is the collection terminal for violation detection, and the second terminal is the remaining collection terminal within the same area as the first terminal. Construct an initial violation prediction model with the first training sample as input, and a reference model with the first training sample and the second training sample as input. The initial violation prediction model and the reference model have the same output layer. Use the first training sample and the second training sample to train the reference model, use the first training sample to train the violation prediction model, and use the target deviation when training the reference model to correct the violation prediction model, and perform iterative training. The specific training process is the same as the training method described above, namely, the first training sample and / or the second training sample are processed by the front-end processing layer to obtain intermediate features, which are then processed by the back-end processing layer to obtain the violation results. By using the target deviation when training the reference model to correct the violation prediction model, the violation prediction model can learn the implicit features of the invisible perspective, thereby being able to make predictions from a panoramic perspective even with input from only a partial explicit perspective, thereby improving the recognition accuracy of violation results.

[0045] In one embodiment, it is understood that the violation prediction model includes multiple processing layers. Therefore, front-end processing layers and back-end processing layers are rationally configured between the multiple processing layers. The various base models of the violation identification model are divided into front-end base models and back-end base models based on demarcation point features. The front-end base models are deployed at the point of entry, and the back-end base models are deployed in the cloud to reduce the data volume of the back-end processing layer in the cloud. Specifically, the data volume output by each processing layer is calculated in order from front to back until the data volume is less than a data volume threshold of the output feature. The processing layer with less than the threshold is then used as the back-end processing layer, and the processing layer with greater than the threshold is used as the front-end processing layer. The data volume threshold of the output feature can be a dynamic threshold, which can be adjusted according to the load status of the output feature, and the base model is transferred between the front-end and back-end processing layers, thereby achieving dynamic deployment of the violation identification model. Specifically, when the load status of the cloud increases, in order to maintain its reduced load status, the base model deployed in the first layer of the back-end processing layer is moved to the front-end. Therefore, this application enables dynamic deployment of multiple processing layers within the violation prediction model to reduce computing pressure in the cloud. Furthermore, through joint training, the initial violation prediction model is decoupled into the front-end and back-end processing layers, which are trained separately. Iterative parameter adjustment ultimately yields a violation prediction model that meets pre-set requirements, improving data processing efficiency and the accuracy of the resulting violation results.

[0046] In one embodiment, step S130: obtaining and outputting the violation result includes: generating warning information according to the violation result and outputting it to a corresponding warning terminal.

[0047] In one embodiment, if the execution terminal of this application is an identification terminal, then the device that receives the warning information is a warning terminal. The two can be the same device, for example, both are remote monitoring devices. Then, after the identification terminal monitors and identifies the violation results, a warning message can be directly generated and output to the monitoring personnel, informing them that a person has violated the law, etc. The violation results can be provided to the violation processing interface for display, so that the business personnel can make decisions and operations on the identified violations. It can be applicable to implementation scenarios including but not limited to traffic safety, industrial production, etc. In other embodiments, the identification terminal and the warning terminal are different devices. For example, in the implementation scenario of industrial production, the identification terminal is used for business personnel to make decisions and operations on the identified violations; and the warning terminal can be a mobile device carried by the operator, which is used to output a warning message in time to inform the operator to stop or correct the violation when the operator has or may violate the law, so as to avoid the occurrence of safety accidents and provide safety protection for the operator. In this embodiment, wireless communication is established between the identification end and the warning end. The wireless communication technology may include but is not limited to: Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), wideband code division multiple access (W-CDMA), code division multiple access (CDMA), time division multiple access (TDMA), Bluetooth, Wireless Fidelity (WiFi) (such as Institute of Electrical and Electronics Engineers standards IEEE 802.11a, IEEE802.11b, IEEE802.11g and / or IEEE 802.11n), Voice over Internet protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging and short messages, and any other suitable communication protocols, even those that have not yet been developed.Furthermore, specific forms of warning terminals include, but are not limited to, mobile phones, tablet computers, personal digital assistants (PDAs), mobile internet devices (MIDs), and wearable devices (e.g., smartwatches). Applications (e.g., apps) on mobile terminals can process and display relevant information and perform corresponding operations, thereby improving usage efficiency. Therefore, by generating warning information based on violation results and outputting it via the corresponding warning terminal, remote monitoring personnel can make timely decisions and take actions regarding violations, improving their management efficiency. Warnings can also be output to corresponding operating personnel, warning them to promptly correct or cease violations, improving safety and preventing accidents or danger.

[0048] Therefore, the present application can obtain the violation results by inputting the on-site images obtained on-site into the recognition model, so that the violation results with higher accuracy can be obtained without the need for monitoring personnel to manually observe the surveillance video in real time, so that business personnel can make decisions and operations based on the violation results identified; and encrypt the on-site images during the process of transmitting them to the violation recognition model to protect the data security of the on-site images and avoid data leakage.

[0049] Example 2

[0050] Figure 2 FIG1 shows an internal structure diagram of a computer device in an embodiment. The computer device can be a terminal or a server. Figure 2 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus. The memory includes a non-volatile storage medium and an internal memory. The non-volatile storage medium of the computer device stores an operating system and may also store a computer program. When the computer program is executed by the processor, the processor can implement the violation behavior identification method. The internal memory may also store a computer program. When the computer program is executed by the processor, the processor can implement the age identification method. It will be understood by those skilled in the art that Figure 2 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0051] Figure 3A schematic diagram of the module structure of a computer device in one embodiment is shown. The computer device can also be divided into the following modules according to the different functions implemented: an acquisition module 101, an encryption module 102, a construction module 103 and an identification module 104. Among them, the acquisition module 101 is used to obtain the scene image; the encryption module 102 is used to encrypt the scene image to obtain the image to be identified; the construction module 103 is used to build a violation recognition model; the identification module 104 is used to input the image to be identified into the violation recognition model to obtain the violation result and output it. Specifically, the functional steps implemented by each module have been described in detail in the violation behavior identification method described in Example 1 of this application. Please refer to the previous text for details and will not be repeated here.

[0052] In one embodiment, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the method described in the first embodiment.

[0053] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0054] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0055] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A method for identifying traffic violation behavior, characterized in that: The steps include: Acquire on-site images; Encrypting the scene image to obtain an image to be identified; Constructing a violation recognition model, inputting the image to be recognized into the violation recognition model to obtain and output a violation result; Before encrypting the scene image to obtain the image to be identified, the method further includes: When acquiring the scene image, acquiring audio information collected at the same time as the image to be identified; Inputting the scene image into a sound prediction model to obtain predicted audio information; determining whether the live image is a real image based on the audio information and the predicted audio information; When the on-site image is a real image, the on-site image is encrypted.

2. The method for identifying traffic violations according to claim 1, wherein: Before encrypting the scene image to obtain the image to be identified, the method further includes: Obtain the identity information of the operator; Obtaining a public key and a private key matching the identity information according to the identity information; The step of encrypting the scene image to obtain the image to be identified includes: Extracting facial features from the scene image, determining the identity information corresponding to the scene image based on the facial features; and matching the corresponding private key based on the identity information; The on-site image is signed according to the private key to obtain an encrypted image, and the encrypted image and the on-site image are packaged as the image to be identified.

3. The method for identifying traffic violation behavior according to claim 2, wherein: The step of inputting the image to be identified into the traffic violation identification model includes: decrypting the image to be identified according to the public key to obtain a decrypted image; When the decrypted image is consistent with the on-site image, the on-site image is input into the violation recognition model.

4. The method for identifying traffic violations according to claim 1, wherein: The step of inputting the image to be identified into the violation identification model to obtain and output a violation result includes: Determining an object to be identified in the image to be identified, and identifying key nodes of the object to be identified; Determining the behavior information of the identified object based on the position relationship and position changes of multiple key nodes; Obtaining a preset illegal action recognition mapping table, querying the illegal action recognition mapping table according to the behavior information to determine whether the behavior information belongs to an illegal behavior, the illegal action recognition mapping table being used to store a mapping relationship between the behavior information and the illegal behavior; When the behavior information is a violation, the corresponding violation result is output, and the violation result includes a violation probability and a violation type attribute.

5. The method for identifying traffic violations according to claim 4, wherein: The determining of the identification object in the image to be identified and identifying the key nodes of the identification object includes: When the key node is blocked, converting the image to be recognized into a panoramic recognition image; Key nodes of the recognition object are recognized according to the panoramic recognition image, where the key nodes include display key nodes and occlusion key nodes.

6. The method for identifying traffic violation behavior according to claim 1, wherein: The construction of the violation identification model includes: Obtain a first training sample collected by a first terminal and a second training sample collected by a second terminal, where the first terminal is a collection terminal for violation detection, and the second terminal is another collection terminal within the same area as the first terminal; Constructing an initial traffic violation recognition model using the first training sample as an input, and a reference model using the first training sample and the second training sample as input, wherein the traffic violation recognition model and the reference model have the same output layer; Using the first training sample to train the initial violation recognition model to obtain a first violation result, and using the first training sample and the second training sample to train the reference model to obtain a second violation result; A target deviation is obtained according to the first violation result and the second violation result, and the initial violation recognition model is iteratively adjusted according to the target deviation until the target deviation is less than a preset deviation threshold, thereby obtaining the violation recognition model.

7. The method for identifying traffic violation behavior according to any one of claims 1 to 6, characterized in that: Obtaining and outputting the violation result includes: A warning message is generated according to the violation result and output to a corresponding warning terminal.

8. A computer device, characterized in that: Including processor and memory: The processor is configured to execute the computer program stored in the memory to implement the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images