Deepseek large model-based vehicle tracking method and system

Abstract

The invention is suitable for the technical field of traffic management, and provides a vehicle tracking method and system based on a Deepseek large model, and the method comprises the steps: obtaining real-time monitoring video data based on an electronic camera, processing the real-time monitoring video data based on a processing algorithm set, generating target data information, and transmitting the target data information to the electronic camera. And uploading the target data information to a preset target database, and finally, in response to a user query instruction, effectively determining related data information based on the target database. According to the method, an advanced DeepSort algorithm can be adopted, an adaptive auxiliary tool is combined, efficient and accurate data extraction can be carried out on the shot video, valuable information in the video is mined out, then deep retrieval and intelligent matching are carried out on related data in the MySQL database through a locally deployed privatized DeepSeek large model, the personalized requirements of users are fully considered, and the user experience is improved. And the retrieval result is ensured to have high accuracy and correlation, so that the matching precision and the practical value of the overall retrieval are remarkably improved.

Description

Technical Field

[0001] This application relates to the technical field of traffic management, and more specifically, to a vehicle tracking method and system based on the Deepseek large model. Background Technology

[0002] With the increasing popularity of automobiles, while bringing convenience to people's lives and production, they have also caused traffic congestion and frequent traffic accidents.

[0003] Currently, the sort algorithm is typically used to track vehicles, combined with the virtual loop algorithm to count traffic flow, and then the data sequence is arranged in a conventional manner. However, during the sorting process, due to the lack of in-depth analysis of content relevance, semantic understanding, and user needs, it is difficult to guarantee accurate matching of search results in data retrieval scenarios, resulting in low accuracy of search result matching, which needs further improvement. Summary of the Invention

[0004] Based on this, this application provides a vehicle tracking method and system based on the deepseek large model to solve the problem of low matching accuracy of retrieval results in the prior art.

[0005] In a first aspect, embodiments of this application provide a vehicle tracking method based on a deepseek large model, the method comprising: Real-time monitoring video data is acquired based on preset electronic cameras; Based on a preset set of processing algorithms, the real-time monitoring video data is processed to generate target data information; Upload the target data information to a preset target database; In response to a user query command, relevant data information is determined based on the target database.

[0006] Compared with existing technologies, the beneficial effects are as follows: The vehicle tracking method based on the deepseek large model provided in this application allows the terminal device to first acquire real-time monitoring video data using an electronic camera, then process the real-time monitoring video data using a set of processing algorithms to generate target data information, then upload the target data information to a preset target database, and finally respond to the user's query command by effectively determining relevant data information based on the target database. This ensures accurate matching of search results according to user needs, improves the matching accuracy of search results, and to a certain extent solves the problem of low matching accuracy of current search results.

[0007] Secondly, embodiments of this application provide a vehicle tracking system based on a deepseek large model, the system comprising: Real-time monitoring video data acquisition module: used to acquire real-time monitoring video data based on preset electronic cameras; Target data information generation module: used to process the real-time monitoring video data based on a preset set of processing algorithms to generate target data information; Target data information upload module: used to upload the target data information to a preset target database; Relevant data information determination module: used to determine relevant data information based on the target database in response to user query instructions.

[0008] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in the first aspect above.

[0009] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect above.

[0010] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0012] Figure 1 This is a schematic flowchart of a vehicle tracking method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating step S200 in a vehicle tracking method provided in an embodiment of this application; Figure 3 This is a schematic diagram of a MySQL database provided in one embodiment of this application; Figure 4 This is a flowchart illustrating step S400 in a vehicle tracking method provided in an embodiment of this application; Figure 5 This is a schematic diagram of relevant data information provided in an embodiment of this application; Figure 6 This is a block diagram of a vehicle tracking system provided in one embodiment of this application; Figure 7 This is a schematic diagram of a terminal device provided in an embodiment of this application. Detailed Implementation

[0013] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0014] In the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0015] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0016] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0017] Please see Figure 1 , Figure 1 This is a flowchart illustrating the vehicle tracking method based on the deepseek large model provided in this application embodiment. In this embodiment, the vehicle tracking method is executed by a terminal device. It is understood that the types of terminal devices include, but are not limited to, tablet computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), etc. This application embodiment does not impose any restrictions on the specific type of terminal device.

[0018] Please see Figure 1 The vehicle tracking method provided in this application includes, but is not limited to, the following steps: In the S100, real-time monitoring video data is acquired based on a preset electronic camera.

[0019] Specifically, the terminal device can use a pre-set electronic camera to acquire real-time monitoring video data. The real-time monitoring video data describes the monitoring video obtained by the electronic camera in real time capturing traffic roads. The real-time monitoring video data includes multiple frames of real-time monitoring image information. The real-time monitoring image information includes at least one target object, which may be a vehicle.

[0020] In S200, real-time monitoring video data is processed based on a preset set of processing algorithms to generate target data information.

[0021] Specifically, after the terminal device acquires real-time monitoring video data, it can process the data based on a preset set of processing algorithms to effectively generate target data information. The processing algorithm set includes the DeepSort algorithm, state equation, observation equation, and joint matching cost function. The target data information is used to describe valuable data.

[0022] In some possible implementations, please refer to [link / reference needed] for generating the target data information. Figure 2 Step S200 includes, but is not limited to, the following steps: In S210, based on the preset DeepSort algorithm, the state feature set information of each target object is obtained for each frame of real-time monitoring image information in the real-time monitoring video data.

[0023] Specifically, the terminal device can perform this processing on each frame of real-time monitoring image information in the real-time monitoring video data based on the preset DeepSort algorithm: obtain the state feature set information of each target object, wherein the state feature set information includes center coordinate information, aspect ratio information, height information, first velocity component information, second velocity component information, and third velocity component information. The first velocity component information corresponds to the horizontal coordinate of the center coordinate information, the second velocity component information corresponds to the vertical coordinate of the center coordinate information, and the third velocity component information corresponds to the height information.

[0024] In S220, the state feature set information is input into the preset Kalman filter motion model to determine the initial state vector information of the target object.

[0025] Specifically, after the terminal device acquires the state feature set information, it can input the state feature set information into the preset Kalman filter motion model to quickly determine the initial state vector information of the target object.

[0026] In one possible implementation, the aforementioned initial state vector information could be: , In the formula, This represents the initial state vector information. The x-coordinate represents the center coordinate information. The ordinate represents the center coordinate information. This indicates the aspect ratio information. Indicates altitude information. This indicates the first velocity component information. This indicates the second velocity component information. This indicates the third velocity component information.

[0027] In S230, the initial state vector information is input into the preset state equation to determine the target state vector information.

[0028] Specifically, after the terminal device determines the initial state vector information, it can input the initial state vector information into a preset state equation to effectively determine the target state vector information. In one possible implementation, the above state equation could be: , In the formula, This represents the target state vector information, that is, the state vector of the target object at time k, which may contain... ; This represents the preset state transition matrix information, which is used to express the motion law of the target object; This indicates the preset input influence control factor information; This represents the preset process noise information, which follows a zero-mean Gaussian distribution.

[0029] In S240, the target state vector information is input into the preset observation equation to determine the observation information.

[0030] Specifically, after the terminal device determines the target state vector information, it can input the target state vector information into a preset observation equation to effectively determine the observation value information. In one possible implementation, the above observation equation could be: , In the formula, This represents the observation information, i.e., the detection box. ; This represents a preset observation matrix, which is used to map the state vector to the measurement space; This represents the preset observation noise information, which follows a Gaussian distribution.

[0031] In S250, trajectory association feature information is determined based on observation information and a preset joint matching cost function.

[0032] Specifically, after the terminal device effectively determines the observation information, it can efficiently determine the trajectory association feature information based on the observation information and the preset joint matching cost function.

[0033] In one possible implementation, the joint matching cost function described above could be: , In the formula, Indicates trajectory association feature information; This indicates the preset weighting parameter information. The value of the weighting parameter information can be 0.98. The weighting parameter information is used to balance the confidence of motion and appearance. This represents the preset Mahalanobis distance information, which is used to represent motion similarity and to calculate the difference between the j-th and i-th detected objects in the trajectory prediction state.

[0034] In one possible implementation, the formula for calculating the Mahalanobis distance information mentioned above could be: , In the formula, This represents the observation information corresponding to the j-th detected object. This represents the observation matrix corresponding to the trajectory of the i-th object. This represents the covariance matrix of the Kalman filter prediction. This shows the observation information corresponding to the i-th detected object.

[0035] In the formula, This represents the preset cosine distance information, which is used to represent appearance similarity. The cosine distance information is obtained through the deep feature vector. The minimum distance between the trajectory feature library and the target feature library.

[0036] In one possible implementation, the formula for calculating the cosine distance information mentioned above could be: , In the formula, For the capacity of the trajectory feature library, The sequence number of the trajectory. This represents the depth feature vector of the j-th detected object. Indicates the first The feature vector corresponding to the trajectory of an object.

[0037] In S260, the optimal matching information is determined based on the preset Hungarian algorithm and trajectory association feature information.

[0038] Specifically, after the terminal device determines the trajectory association feature information, it can determine the optimal matching information based on the preset Hungarian algorithm and trajectory association feature information, thereby achieving detection and trajectory association by minimizing the total matching cost.

[0039] In one possible implementation, the above Hungarian algorithm could be: , In the formula, This represents the optimal matching information, i.e., all possible matching combinations; This represents the cost element of the input; it should be noted that the constraint of this Hungarian algorithm is that each detection matches at most one trajectory, and vice versa.

[0040] In S270, the optimal matching information is determined to be the target data information.

[0041] Specifically, after the terminal device determines the optimal matching information, the terminal device can determine the optimal matching information as the target data information.

[0042] In some possible implementations, in order to improve the accuracy of search result matching, after step S260, the method may include, but is not limited to, the following steps: In S261, based on the optimal matching information, the Kalman filter state corresponding to the Kalman filter motion model is updated, and trajectory association feature information without matching is retained or deleted.

[0043] Specifically, the terminal device can update the Kalman filter state corresponding to the Kalman filter motion model based on the optimal matching information, thereby updating the Kalman filter state with successfully matched detections; at the same time, the terminal device can retain or delete the trajectory association feature information that has not been matched, thereby achieving temporary retention or deletion of unmatched trajectories.

[0044] In some possible implementations, to further improve the accuracy of search result matching, after step S270, the method may include, but is not limited to, the following steps: In S271, the target data information is filtered based on a preset chi-square test mechanism to determine the filtered target data information.

[0045] Specifically, the terminal device can filter the target data information based on a preset chi-square test mechanism to determine the filtered target data information.

[0046] In one possible implementation, the chi-square test mechanism described above could be: , In the formula, This represents the chi-square distribution threshold with 4 degrees of freedom; the degrees of freedom for the chi-square distribution threshold correspond to a four-dimensional measurement. When the value of is 0.95, the chi-square distribution threshold is approximately 9.49, which can be taken as 9.49. In S300, the target data information is uploaded to the preset target database.

[0047] Specifically, for example, please refer to Figure 3 After the terminal device generates valuable data, it can upload the target data information to a preset target database, which may be a MySQL database.

[0048] In S400, in response to a user query command, relevant data information is determined based on the target database.

[0049] Specifically, after the terminal device uploads the target data information, the terminal device can respond to the user's query command, determine the relevant data information based on the target database, thereby ensuring accurate matching of search results according to user needs and improving the matching accuracy of search results. Among them, the deepseek large model includes the langchain module, which includes the GROP algorithm, vector similarity calculation function, joint generation probability model and standard autoregressive model.

[0050] In some possible implementations, to determine the relevant data information, please refer to [link / reference]. Figure 4 Step S400 includes, but is not limited to, the following steps: In S410, in response to a user query command, the optimization target information is determined based on the GROP algorithm.

[0051] Specifically, the terminal device can respond to user query commands and determine optimization target information based on the GROP algorithm with a unified training objective, thereby enabling the deepseek large model to unify the optimization objectives of supervised fine-tuning (SFT) and reinforcement learning (RL) through the GRPO algorithm.

[0052] In one possible implementation, the above GROP algorithm can be: , In the formula, This indicates the target information to be optimized. Indicates the expectation symbol, Indicates output From the old strategy generate, This represents the probability distribution of the model generating output y for input x. This represents a preset reward function, which can be designed based on human feedback or rules. This represents the preset intensity parameter, which is used to balance the intensity of exploration and utilization.

[0053] In S420, vector similarity information is determined based on the vector similarity calculation function and the optimization target information.

[0054] Specifically, after the terminal device determines the optimization target information, it can determine the vector similarity information based on the optimization target information using a vector similarity calculation function. This enables the fusion of knowledge base retrieval and generation through the following mathematical methods when integrating DeepSeek-R into Langchain.

[0055] In one possible implementation, the vector similarity calculation function described above could be: , In the formula, Represents vector similarity information. This indicates a query for the embedding vector information corresponding to q. This represents the embedding vector information corresponding to document d. It should be noted that highly similar documents are prioritized for answer generation to improve real-time accuracy.

[0056] In S430, based on the joint generation probability model, the relevance probability information is determined according to the vector similarity information.

[0057] Specifically, after the terminal device determines the vector similarity information, it can determine the relevance probability information based on the vector similarity information using a joint generation probability model.

[0058] In one possible implementation, the above joint generation probability model could be: , In the formula, This indicates the probability of correlation. This represents the probability of relevance between document d and input x. This represents the conditional generation probability based on document d.

[0059] In S440, the target probability information is determined based on the standard autoregressive model and the correlation probability information.

[0060] Specifically, since DeepSeek-R1's text generation follows a standard autoregressive model, after the terminal device determines the relevance probability information, it can determine the target probability information based on the relevance probability information according to the standard autoregressive model.

[0061] In one possible implementation, the above standard autoregressive model could be: , In the formula, This indicates the probability of correlation. Indicates the sequence number of the steps. Indicates the total number of steps. Indicates the first The token generated in each step Indicates the first -1 The token generated at the first point in time in the first step.

[0062] In one possible implementation, the terminal device can dynamically compute the context dependencies using the Transformer's attention mechanism, i.e.:

[0063] In the formula, Q, K, and V represent the query, key, and value matrices, respectively. This is the dimension scaling factor.

[0064] In S450, relevant data information is determined based on target probability information and preset core component information.

[0065] Specifically, after the terminal device determines the target probability information, it can effectively determine relevant data information based on the target probability information and the preset core component information, where the core component information is the LLMChain component.

[0066] For example, a terminal device can abstract the chain structure in LangChain (such as LLMChain) into a chain decomposition of conditional probabilities, thereby representing multi-step reasoning and contextual coherence through the chain structure, i.e.: , In the formula, This represents the logical combination of the model output and the prompt.

[0067] For example, please refer to Figure 5 , Figure 5 This demonstrates how to connect a large DeepSeek-R1 model to a MySQL database via LangChain, which is deployed locally and privately, and then display the results when querying the large model through the WEBUI.

[0068] It should be noted that this application employs the more advanced DeepSort algorithm. Multi-target tracking is achieved by combining the basic structure of DeepSort with a Kalman filter motion model and a deep feature appearance model. Tools used include imutils from OpenCV, the DNN deep learning module, the IOU (Intersection over Union) function, the Python module FilterPy for filter implementation, and the YOLOv3 model. Real-time video data is collected from electronic cameras on highways, and the data is analyzed, cleaned, and filtered using the above algorithms and tools to obtain valuable data, which is then stored in a MySQL database. OpenCV's main functions are image processing, video analysis, and object detection. Its image processing module performs image filtering, and its video analysis module mainly performs video stream capture, motion tracking, background modeling, and optical flow analysis, using the Kalman filter algorithm. Imutils is a lightweight wrapper library for OpenCV, providing high-level APIs to simplify image processing. Its main functions include image resizing (i.e., maintaining aspect ratio), rotation, translation, edge detection, skeletonization, brightness / contrast adjustment, Matplotlib image display optimization, contour sorting, and perspective transformation. Meanwhile, FileterPy is a Python library focused on state estimation and filtering algorithms, supporting core algorithms such as the Kalman Filter (KF) and Extended Kalman Filter (EKF), and is suitable for scenarios such as signal processing, target tracking, and sensor fusion. YOLOv3 is a landmark single-stage object detection model in the YOLO series, with real-time detection speed and multi-scale prediction capabilities as its core advantages, and is suitable for acquiring traffic flow and speed data. This application also deploys a locally private DeepSeek large model, using the langchain module in DeepSeek to connect to a MySQL database for vehicle tracking and traffic flow data, and uses a dialog-based approach to obtain real-time vehicle data that meets specific conditions during traffic movement.

[0069] The implementation principle of the vehicle tracking method based on the DeepSeek large model in this application embodiment is as follows: The terminal device can first acquire real-time monitoring video data based on an electronic camera, then process the real-time monitoring video data based on a set of processing algorithms to generate target data information, then upload the target data information to a preset target database, and finally respond to the user's query command, effectively determine relevant data information based on the target database, thereby achieving accurate matching of search results according to user needs and improving the matching accuracy of search results.

[0070] It should be noted that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0071] Embodiments of this application also provide a vehicle tracking system based on a deepseek large model. For ease of explanation, only the parts relevant to this application are shown, such as... Figure 6 As shown, the system 60 includes: Real-time monitoring video data acquisition module 61: used to acquire real-time monitoring video data based on a preset electronic camera; Target data information generation module 62: Used to process real-time monitoring video data based on a preset set of processing algorithms to generate target data information; Target data information upload module 63: used to upload target data information to a preset target database; Relevant data information determination module 64: In response to user query commands, it determines relevant data information based on the target database.

[0072] Optionally, the real-time monitoring video data includes multiple frames of real-time monitoring image information, the real-time monitoring image information includes at least one target object, and the processing algorithm set includes the DeepSort algorithm, state equation, observation equation, and joint matching cost function; the aforementioned target data information generation module 62 includes: The state feature set information acquisition submodule is used to acquire the state feature set information of each target object based on the preset DeepSort algorithm for each frame of real-time monitoring image information in real-time monitoring video data. The state feature set information includes center coordinate information, aspect ratio information, height information, first velocity component information, second velocity component information, and third velocity component information. The first velocity component information corresponds to the horizontal coordinate of the center coordinate information, the second velocity component information corresponds to the vertical coordinate of the center coordinate information, and the third velocity component information corresponds to the height information. Initial state vector information determination submodule: used to input the state feature set information into the preset Kalman filter motion model to determine the initial state vector information of the target object; Target state vector information determination submodule: Used to input the initial state vector information into the preset state equation to determine the target state vector information; The observation information determination submodule is used to input the target state vector information into the preset observation equation to determine the observation information. The trajectory association feature information determination submodule is used to determine trajectory association feature information based on observation information and a preset joint matching cost function. The optimal matching information determination submodule is used to determine the optimal matching information based on the preset Hungarian algorithm and trajectory association feature information. The target data information determination submodule is used to determine the optimal matching information as the target data information, which describes valuable data.

[0073] Optionally, the deepseek large model includes the langchain module, which includes the GROP algorithm, vector similarity calculation function, joint generation probability model, and standard autoregressive model; the aforementioned related data information determination module 64 includes: The optimization target information determination submodule is used to determine the optimization target information based on the GROP algorithm in response to user query commands. Vector similarity information determination submodule: used to determine vector similarity information based on the vector similarity calculation function and the optimization target information; The correlation probability information determination submodule is used to determine the correlation probability information based on vector similarity information and the joint generation probability model. Target probability information determination submodule: used to determine target probability information based on the standard autoregressive model and the correlation probability information; The relevant data information determination submodule is used to determine relevant data information based on the target probability information and the preset core component information, where the core component information is the LLMChain component.

[0074] Optionally, the system 60 also includes: Update processing module: It is used to update the Kalman filter state corresponding to the Kalman filter motion model based on the optimal matching information, and to retain or delete the trajectory association feature information that has no matching.

[0075] Optionally, the system 60 also includes: Target data information determination module: This module is used to filter target data information based on a preset chi-square test mechanism to determine the filtered target data information.

[0076] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0077] This application also provides a terminal device, such as... Figure 7 As shown, the terminal device 70 of this embodiment includes: a processor 71, a memory 72, and a computer program 73 stored in the memory 72 and executable on the processor 71. When the processor 71 executes the computer program 73, it implements the steps described in the vehicle tracking method embodiment above, for example... Figure 1 Steps S100 to S400 are shown; or, when processor 71 executes computer program 73, it implements the functions of each module in the above-described device, for example... Figure 6 The functions of modules 61 to 64 are shown.

[0078] The terminal device 70 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device, and includes, but is not limited to, a processor 71 and a memory 72. Those skilled in the art will understand that... Figure 7 This is merely an example of terminal device 70 and does not constitute a limitation on terminal device 70. It may include more or fewer components than shown, or combine certain components, or different components. For example, terminal device 70 may also include input / output devices, network access devices, buses, etc.

[0079] The processor 71 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.; the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0080] The memory 72 can be an internal storage unit of the terminal device 70, such as the hard disk or memory of the terminal device 70. The memory 72 can also be an external storage device of the terminal device 70, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the terminal device 70. Furthermore, the memory 72 can include both internal storage units and external storage devices of the terminal device 70. The memory 72 can also store computer program 73 and other programs and data required by the terminal device 70. The memory 72 can also be used to temporarily store data that has been output or will be output.

[0081] One embodiment of this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0082] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the methods, principles and structures of this application should be covered within the scope of protection of this application.

Claims

1. A vehicle tracking method based on a deepseek large model, characterized in that, The method includes: Real-time monitoring video data is acquired based on preset electronic cameras; Based on a preset set of processing algorithms, the real-time monitoring video data is processed to generate target data information; Upload the target data information to a preset target database; In response to a user query command, relevant data information is determined based on the target database.

2. The method according to claim 1, characterized in that, The real-time monitoring video data includes multiple frames of real-time monitoring image information, and the real-time monitoring image information includes at least one target object. The processing algorithm set includes the DeepSort algorithm, state equation, observation equation, and joint matching cost function. The real-time monitoring video data is processed based on the preset processing algorithm set to generate target data information, including: Based on the preset DeepSort algorithm, for each frame of real-time monitoring image information in the real-time monitoring video data, the state feature set information of each target object is obtained, wherein the state feature set information includes center coordinate information, aspect ratio information, height information, first velocity component information, second velocity component information, and third velocity component information. The first velocity component information corresponds to the horizontal coordinate of the center coordinate information, the second velocity component information corresponds to the vertical coordinate of the center coordinate information, and the third velocity component information corresponds to the height information. The state feature set information is input into a preset Kalman filter motion model to determine the initial state vector information of the target object, wherein the initial state vector information is: ， In the formula, This refers to the initial state vector information. The x-coordinate of the center coordinate information is... The ordinate of the center coordinate information is... The aspect ratio information, For the height information, This is the first velocity component information. This is the second velocity component information. This refers to the third velocity component information; The initial state vector information is input into a preset state equation to determine the target state vector information, wherein the state equation is: ， In the formula, This refers to the target state vector information. This is the preset state transition matrix information. The preset input influence control factor information, This is the preset process noise information; The target state vector information is input into a preset observation equation to determine the observation value information, wherein the observation equation is: ， In the formula, The observed information, For the preset observation matrix, This is the preset observation noise information; Based on the observed information and a preset joint matching cost function, trajectory association feature information is determined, wherein the joint matching cost function is: , In the formula, For the trajectory associated feature information, For the preset weight parameter information, The information is the preset Mahalanobis distance. This is the preset cosine distance information; Based on the preset Hungarian algorithm and the trajectory association feature information, the optimal matching information is determined, wherein the Hungarian algorithm is: , In the formula, This is the optimal matching information. The cost element is the input; The optimal matching information is determined as target data information, wherein the target data information is used to describe valuable data.

3. The method according to claim 1, characterized in that, The DeepSeek large model includes a LangChain module, which comprises the GROP algorithm, a vector similarity calculation function, a joint generation probability model, and a standard autoregressive model. In response to a user query command, based on the target database, relevant data information is determined, including: In response to a user query command, optimization target information is determined based on the GROP algorithm, wherein the GROP algorithm is: ， In the formula, For the optimization target information, For the expected symbol, Indicates output From the old strategy generate, This represents the probability distribution for the model to generate output y from input x. For the preset reward function, These are preset strength parameters; Based on the vector similarity calculation function, and according to the optimization target information, vector similarity information is determined, wherein the vector similarity calculation function is: , In the formula, The vector similarity information, To query the embedding vector information corresponding to q, This refers to the embedding vector information corresponding to document d; Based on the joint generation probability model, and according to the vector similarity information, the relevance probability information is determined, wherein the joint generation probability model is: ， In the formula, This refers to the correlation probability information. Let be the correlation probability between document d and input x. The conditional generation probability is based on document d; Based on the standard autoregressive model, the target probability information is determined according to the correlation probability information, wherein the standard autoregressive model is: ， In the formula, This refers to the correlation probability information. The step number is used to indicate the sequence of steps. The total number of steps. For the first The token generated in each step For the first -1 step to generate the token at the first point in time; Based on the target probability information, relevant data information is determined according to the preset core component information, wherein the core component information is the LLMChain component.

4. The method according to claim 2, characterized in that, After determining the optimal matching information based on the preset Hungarian algorithm and the trajectory association feature information, the method further includes: Based on the optimal matching information, the Kalman filter state corresponding to the Kalman filter motion model is updated, and trajectory association feature information that does not match is either retained or deleted.

5. The method according to claim 2, characterized in that, After determining the optimal matching information as the target data information, the method further includes: Based on a preset chi-square test mechanism, the target data information is filtered to determine the filtered target data information.

6. A vehicle tracking method system based on a deepseek large model, characterized in that, The system includes: Real-time monitoring video data acquisition module: used to acquire real-time monitoring video data based on preset electronic cameras; Target data information generation module: used to process the real-time monitoring video data based on a preset set of processing algorithms to generate target data information; Target data information upload module: used to upload the target data information to a preset target database; Relevant data information determination module: used to determine relevant data information based on the target database in response to user query instructions.

7. The system according to claim 6, characterized in that, The system also includes: Target data information determination module: used to filter the target data information based on a preset chi-square test mechanism, and determine the filtered target data information.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 5.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 5.

Images