Pedestrian trace acquisition method, apparatus and device, and storage medium

Abstract

The invention discloses a pedestrian trace acquisition method and device, equipment and a storage medium, belongs to the field of artificial intelligence, and is used for accurately acquiring a pedestrian trace. The method comprises the steps that multiple pieces of image data of a first picture at a first moment are acquired, feature fusion is carried out on the image data based on a preset weight coefficient, a first fused image is obtained, and the image data comprise at least two of a visible light image, an infrared image and a code detection image; determining the first state information of the first pedestrian in the first fused image, and predicting the state information of the next moment based on the first state information to obtain predicted state information; acquiring a plurality of image data of the first picture at the next moment, and performing feature fusion on the image data based on a preset weight coefficient to obtain a second fused image; and determining second state information of the first pedestrian in the second fused image, and determining a pedestrian trace of the first pedestrian based on the second state information and the predicted state information.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence, and specifically relates to a method, device, equipment and storage medium for obtaining pedestrian traces. Background Technology

[0002] Pedestrian tracking has long been one of the most important research areas in computer vision. Due to its high practical value and applicability, it has attracted significant attention from researchers across various fields, with a large number of scientists joining its research community. In recent years, multimodal fusion techniques combining infrared and visible light images have gradually become a research hotspot in pedestrian tracking. By fusing features from two different modalities, the aim is to improve the accuracy and stability of pedestrian recognition in complex scenes. However, in certain environments, such as dark environments, the quality of visible light images can be affected, and in high-temperature environments or under strong infrared background interference, the quality of infrared images may be affected, leading to a decrease in the accuracy of the acquired pedestrian tracks.

[0003] Therefore, a method for accurately obtaining pedestrian tracks is needed. Summary of the Invention

[0004] This application provides a method for obtaining pedestrian tracks, which can accurately obtain pedestrian tracks.

[0005] In a first aspect, embodiments of this application provide a method for obtaining pedestrian tracks. The method includes: acquiring multiple image data of a first screen at a first moment, and fusing them based on preset weight coefficients to obtain a first fused image, wherein the first screen is the screen for which pedestrian tracks are to be obtained, and the image data includes at least two of visible light images, infrared images, and detection images; determining first state information of a first pedestrian in the first fused image, and predicting state information for a next moment based on the first state information to obtain predicted state information, wherein the first pedestrian is the pedestrian for which pedestrian tracks are to be obtained in the first fused image, and the first state information is the state information of the first pedestrian in the fused image; acquiring multiple image data of the first screen at the next moment, and fusing them based on the preset weight coefficients to obtain a second fused image; determining second state information of the first pedestrian in the second fused image, and determining the pedestrian track of the first pedestrian based on the second state information and the predicted state information, wherein the second state information is the state information of the first pedestrian in the second fused image.

[0006] Secondly, embodiments of this application provide a pedestrian tracking acquisition device, comprising: a first acquisition module, configured to acquire multiple image data of a first screen at a first moment, and fuse them based on a preset weight coefficient to obtain a first fused image, wherein the first screen is the screen for which pedestrian tracking is to be acquired, and the image data includes at least two of visible light images, infrared images, and detection images; a first determination module, configured to determine first state information of a first pedestrian in the first fused image, and predict state information for a next moment based on the first state information to obtain predicted state information, wherein the first pedestrian is the pedestrian for which pedestrian tracking is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image; a second acquisition module, configured to acquire multiple image data of the first screen at the next moment, and fuse them based on the preset weight coefficient to obtain a second fused image; and a second determination module, configured to determine second state information of the first pedestrian in the second fused image, and determine the pedestrian tracking of the first pedestrian based on the second state information and the predicted state information, wherein the second state information is the state information of the first pedestrian in the second fused image.

[0007] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps of the method described in the first aspect.

[0008] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0009] Fifthly, embodiments of this application provide a computer program product that, when executed by a processor, implements the steps of the method described in the first aspect.

[0010] In a sixth aspect, embodiments of this application provide a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the method as described in the first aspect.

[0011] In this embodiment, multiple image data of a first screen at a first moment are acquired and their features are fused based on a preset weight coefficient to obtain a first fused image. The first screen is the screen for which pedestrian tracking is to be acquired, and the image data includes at least two of visible light images, infrared images, and detection images. The first state information of a first pedestrian in the first fused image is determined, and the state information of the next moment is predicted based on the first state information to obtain predicted state information. The first pedestrian is the pedestrian for which pedestrian tracking is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image. Multiple image data of the first screen at the next moment are acquired and their features are fused based on a preset weight coefficient to obtain a second fused image. The second state information of the first pedestrian in the second fused image is determined, and the pedestrian tracking of the first pedestrian is determined based on the second state information and the predicted state information. The second state information is the state information of the first pedestrian in the second fused image, which can accurately acquire pedestrian tracking. Attached Figure Description

[0012] Figure 1 This is a flowchart illustrating a method for obtaining pedestrian tracks provided in an embodiment of this application; Figure 2 This is a flowchart illustrating the second method for obtaining pedestrian traces provided in this application embodiment; Figure 3 This is a flowchart illustrating the third method for obtaining pedestrian traces provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of a pedestrian tracking device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a pedestrian tracking device provided in an embodiment of this application. Detailed Implementation

[0013] 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 part of the embodiments of this application, not all of them. 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.

[0014] The terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of this application can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects; for example, the first object can be one or more. In addition, the term "and / or" in the specification and claims refers to at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.

[0015] The pedestrian tracking method provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0016] Figure 1 This illustration shows a method for obtaining pedestrian tracks according to an embodiment of the present invention. The method can be executed by an electronic device, which may include a server and / or a terminal device, wherein the terminal device may be, for example, a vehicle-mounted terminal or a mobile phone terminal. In other words, the method can be executed by software or hardware installed in a pedestrian track obtaining device, and the method includes the following steps: Step 102: Acquire multiple image data of the first frame at the first moment, and fuse them based on preset weight coefficients to obtain the first fused image.

[0017] The first image is the image from which pedestrian traces are to be acquired, and the image data includes at least two of the following: visible light image, infrared image, and detection image.

[0018] The execution entity of the pedestrian tracking acquisition method described in this application can be a pedestrian tracking acquisition system or pedestrian tracking acquisition software. This embodiment of the application takes a pedestrian tracking acquisition system as an example.

[0019] The pedestrian tracking system acquires multiple image data points of a first frame at a first moment. The first frame is the frame from which pedestrian tracking needs to be acquired, and the image data includes at least two of the following: visible light images, infrared images, and code-detection images. Preferably, the multiple image data points for pedestrian tracking acquisition include all three: visible light images, infrared images, and code-detection images. In other words, at the first moment, the pedestrian tracking system can acquire a visible light image of the first frame using a visible light image acquisition device, an infrared image of the first frame using an infrared image acquisition device, and a code-detection image of the first frame using a code-detection image acquisition device.

[0020] Furthermore, since the image data acquired by the visible light image acquisition device, the infrared image acquisition device, and the code detection image acquisition device are in different formats and sizes, after acquiring multiple image data, the pedestrian tracking acquisition system can also normalize the visible light image, infrared image, and code detection image, unifying them into a preset format, thereby facilitating subsequent processing of multiple image data.

[0021] After acquiring multiple image data, the pedestrian tracking system performs feature fusion on the multiple image data based on preset weight coefficients to obtain a first fused image. The preset weight coefficients include the weights corresponding to each image data when performing feature fusion. For example, the weight coefficients for visible light images, infrared images, and detection images when performing fusion are 0.4, 0.4, and 0.2, respectively.

[0022] Furthermore, when fusing features from multiple image datasets, the pedestrian tracking system can directly fuse the feature information of each image dataset, and it can also acquire and fuse high-level features from each image dataset. High-level features are used to characterize the features that represent the image dataset, or the main features of the image. When acquiring and fusing high-level features from each image dataset, the pedestrian tracking system first acquires the feature vectors of each image dataset, and then fuses the feature vectors based on the weight coefficients corresponding to each image dataset. For example, it can extract local and global features from visible light images, infrared images, and code-detection images using a deep convolutional neural network (CNN) to determine the high-level features corresponding to each type of image, and then fuse these determined high-level features.

[0023] Specifically, when performing feature fusion on various image data, the system can simultaneously perform feature fusion on visible light images, infrared images, and detection images. Alternatively, it can first perform feature fusion on visible light images and infrared images, and then perform feature fusion with the detection image based on the feature fusion result to obtain the first fused image. More specifically, when the pedestrian tracking acquisition system performs feature fusion, it can not only perform a weighted summation of the feature vectors of the visible light image and the infrared image using the weighted summation formula F=W1*A+W2*B, where F is the fused feature vector, A is the feature vector of the visible light image, B is the feature vector of the infrared image, W1 is the weight coefficient corresponding to the visible light image, and W2 is the weight coefficient corresponding to the infrared image, but it can also perform a concatenation summation using the concatenation formula F=[A;B], where ";" represents the vector concatenation operation.

[0024] Step 104: Determine the first state information of the first pedestrian in the first fused image, and predict the state information of the next moment based on the first state information to obtain the predicted state information.

[0025] Wherein, the first pedestrian is the pedestrian whose trace is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image.

[0026] After determining the first fused image, the pedestrian tracking system identifies the first pedestrian in the first fused image and then determines the first pedestrian's first state information. The first pedestrian is the pedestrian in the first fused image whose tracking needs to be performed; it can be any pedestrian in the first fused image whose tracking needs to be performed. The first state information includes the first pedestrian's status information, such as location information, appearance features, speed, and motion state. In other words, the pedestrian tracking system identifies the pedestrian in the first fused image whose tracking needs to be performed as the first pedestrian and then obtains the first pedestrian's first state information.

[0027] After determining the first pedestrian's initial state information, the pedestrian tracking system predicts the pedestrian's state information for the next moment based on this initial state information, thus obtaining the predicted state information. Specifically, the initial state information includes the pedestrian's location, speed, and position. Therefore, the pedestrian tracking system can predict the pedestrian's position and speed for the next moment based on their location, speed, and motion state, thereby obtaining the predicted state information.

[0028] Step 106: Obtain multiple image data of the first frame at the next moment, and fuse them based on the preset weight coefficient to obtain a second fused image.

[0029] After predicting the state information of the first pedestrian at the next moment, the pedestrian tracking system acquires multiple image data of the first scene at the next moment, and performs feature fusion on the acquired multiple image data based on preset weight coefficients to obtain the fused second image. Here, the next moment is the moment after the first moment, and the preset weight coefficients are the same as or different from the weight coefficients when performing feature fusion on the multiple image data of the first moment.

[0030] The first pedestrian and background image in the first frame are constantly moving. The pedestrian tracking system acquires multiple image data from the first frame at two consecutive moments (the first moment and the next moment), and the multiple image data acquired at the first moment and the next moment are not the same. The pedestrian tracking system then continuously acquires multiple image data from the first frame at each moment and fuses the image data to obtain the fused image corresponding to each moment.

[0031] Specifically, the pedestrian tracking system acquires multiple image data of the first frame at the next moment, which are at least two of visible light images, infrared images, and code-detection images. Preferably, the multiple image data of the next moment include visible light images, infrared images, and code-detection images. The method for feature fusion of the multiple image data of the next moment by the pedestrian tracking system is the same as the method for feature fusion of the multiple image data of the first moment described above, and will not be elaborated here.

[0032] Step 108: Determine the second state information of the first pedestrian in the second fused image, and determine the pedestrian trace of the first pedestrian based on the second state information and the predicted state information.

[0033] The second state information refers to the state information of the first pedestrian in the second fused image.

[0034] After acquiring the second fused image at the next moment, the pedestrian tracking system determines the second state information of the first pedestrian in the second fused image. This second state information refers to the state information of the first pedestrian in the second fused image. In other words, after acquiring fused images at various moments in real time, the pedestrian tracking system also determines the state information of the first pedestrian in the fused images at each moment in real time.

[0035] After determining the state information of the first pedestrian in the second fused image, i.e., the second state information, the pedestrian tracking acquisition system determines the pedestrian's track based on the second state information and the predicted state information, and outputs the acquired pedestrian track. The output of the pedestrian track mainly includes the current state information of the first pedestrian (such as speed, position, trajectory information, and predicted state information for the next moment), and this data will be provided in real time to the monitoring system or backend server for subsequent processing, storage, or alarm use.

[0036] The pedestrian tracking method provided in this invention acquires multiple image data of a first screen at a first moment, and fuses them based on preset weight coefficients to obtain a first fused image. The first screen is the screen for which pedestrian tracking is to be acquired, and the image data includes at least two of visible light images, infrared images, and detection images. The method determines the first state information of a first pedestrian in the first fused image, and predicts the state information for the next moment based on the first state information to obtain predicted state information. The first pedestrian is the pedestrian in the first fused image for which pedestrian tracking is to be acquired, and the first state information is the state information of the first pedestrian in the fused image. The method then acquires multiple image data of the first screen at the next moment, and fuses them based on preset weight coefficients to obtain a second fused image. Finally, the method determines the second state information of the first pedestrian in the second fused image, and determines the pedestrian tracking of the first pedestrian based on the second state information and the predicted state information. The second state information is the state information of the first pedestrian in the second fused image, enabling accurate acquisition of pedestrian tracking.

[0037] In one implementation, the step of acquiring multiple image data of the first frame at a first moment and fusing them based on preset weight coefficients to obtain a first fused image (step 102) can be performed via steps A1-A2: Step A1: Obtain the first environmental features of the first scene at the first moment.

[0038] The first environmental feature is used to characterize the environmental features in multiple image data at the first time point.

[0039] After acquiring multiple image data of the first scene at the first moment, the pedestrian tracking system can obtain the first environmental features of the first moment through the multiple image data. The first environmental features are used to characterize the environmental features in the multiple image data at the first moment. For example, environmental features include light intensity, etc.

[0040] Step A2: Determine the first weight coefficient corresponding to the first environmental feature based on the preset weight coefficient correspondence table, and perform feature fusion on multiple image data at the first time based on the first weight coefficient.

[0041] The weight coefficient correspondence table pre-stores the weight coefficients corresponding to different environmental features.

[0042] After determining the first environmental feature at the first moment, the pedestrian tracking system determines the first weight coefficient corresponding to the first environmental feature from a preset weight coefficient correspondence table based on the first environmental feature, and then performs feature fusion on multiple image data at the first moment according to the first weight coefficient. The weight coefficient correspondence table stores the weight coefficients corresponding to different environmental features in advance.

[0043] In other words, the weight coefficient correspondence table pre-stores the weight coefficients corresponding to different environmental features. In this way, when the pedestrian tracking system determines the environmental features at the first moment, it can determine the weight coefficients for feature fusion of multiple image data based on the environmental features.

[0044] Furthermore, after acquiring multiple image data points at the next time step, the pedestrian tracking system can also determine the weighting coefficients for feature fusion of these images based on the data from the next time step. In other words, after acquiring multiple image data points at each time step, the pedestrian tracking system can determine the environmental features at each time step, and then determine the weighting coefficients for each time step using a weighting coefficient mapping table. Finally, feature fusion is performed on the image data from each time step based on these weighting coefficients.

[0045] In this embodiment, by dynamically adjusting the weight coefficients of visible light images, infrared images, and code-detection images, the pedestrian tracking system can flexibly adjust the weight coefficients of each image according to actual conditions (such as nighttime, daytime, or special events), thereby achieving optimal tracking performance in different environments. In the multimodal feature fusion stage, not only are infrared images, visible light images, and code-detection images combined and fused into a unified feature vector through weighted summation or concatenation, but a dynamic weight coefficient adjustment strategy also allows the features of each modality to adaptively adjust according to environmental conditions, maximizing the utilization of information from each modality. These processes enable the pedestrian tracking system to maintain high accuracy and robustness in complex and dynamic environments.

[0046] In one implementation, the step of predicting the state information for the next moment based on the first state information to obtain predicted state information (step 104) can be performed by step B1: Step B1: Based on the first state information, predict the state information of the first pedestrian at the next moment using a pre-trained state prediction model to obtain the predicted state information.

[0047] When predicting the state information of the next moment based on the first state information at the first moment, the pedestrian tracking system can make predictions using a pre-trained state prediction model. This involves inputting the first state information into the pre-trained model to predict the state information of the next moment, thus determining the predicted state information. When predicting the state information of the next moment based on the state prediction model, the input information can include not only the first pedestrian's state information at the first moment, but also the first pedestrian's state information at historical moments and at the second moment (i.e., using the first pedestrian's state information at multiple historical moments as input), so that the state prediction model can better predict the first pedestrian's state information of the next moment.

[0048] Specifically, pedestrian tracking systems can use deep neural networks (such as a combination of CNN and LSTM networks) to predict the next location of a pedestrian in a fused image based on the appearance features and motion trajectory of the first pedestrian. Deep learning can simultaneously consider spatial features (image information) and temporal features (motion trajectory), thus providing more accurate predictions. This is especially true in scenarios where traditional filtering methods struggle to match changes in motion speed and erratic trajectories, where deep learning methods offer better performance.

[0049] Furthermore, the pedestrian tracking system can input the state information of the first pedestrian at the first moment into a deep neural network, enabling the deep neural network to predict the state information of the first pedestrian at the next moment and output the state information of the next moment. This state information includes motion information such as position, velocity, and acceleration. The Long Short-Term Memory (LSTM) network in the deep neural network can effectively capture the motion pattern of the first pedestrian, especially when there are accelerations, decelerations, or changes in turning signals during the pedestrian's movement. LSTM can predict the future trajectory of the target based on historical data. This gives LSTM a strong temporal modeling capability when processing target motion, providing more accurate predictions than traditional Kalman filtering.

[0050] After determining the second state information of the first pedestrian in the second fused image (step 108), step B2 can also be performed: Step B2: Update the parameters of the pre-trained state prediction model based on the predicted state information and the second state information.

[0051] After predicting the state information of the first pedestrian at the next moment through the pre-trained state prediction model and obtaining the predicted state information, the pedestrian tracking system can also update the parameters of the pre-trained state prediction model based on the predicted state information and the second state information.

[0052] Specifically, the predicted state information is the result determined by the pre-trained state prediction model to predict the state information of the first pedestrian at the next moment, while the second state information is the actual state information of the first pedestrian at the next moment. Therefore, the pedestrian tracking acquisition system can update the parameters of the pre-trained state prediction model through the predicted state information and the second state information. That is, it determines the difference between the predicted state information of the pre-trained state prediction model to predict the state of the first pedestrian at the next moment and the actual state information of the first pedestrian at the next moment (i.e., the second state information), and updates the parameters of the pre-trained state prediction model according to the difference so that the state prediction model can better predict the state information of the pedestrian at the next moment.

[0053] In practical applications, pedestrian tracking systems need to possess real-time feedback and adaptive optimization capabilities to ensure that the system can adjust parameters and optimize performance promptly based on the current tracking status. Through real-time feedback, the system can quickly respond to scene changes, adjust algorithm parameters, and optimize based on real-time data, ensuring stability and accuracy over long-term operation. To adjust and optimize the tracking process in real time, the tracking results need to be evaluated in real time. This evaluation process mainly judges the tracking quality by comparing the results with the actual target location or by measuring the prediction error of the tracked target. For example, the predicted position information (or predicted state information) at the first moment and the actual observed position information (or second state information) of the first pedestrian at the next moment are determined. Then, the tracking error is determined using Euclidean distance. Finally, when the error exceeds a preset threshold, the parameters of the state prediction model are adjusted or the algorithm is optimized.

[0054] In one implementation, steps C1-C3 can also be performed: Step C1: Obtain the first environmental features of the first screen at the first moment and the second environmental features at the next moment.

[0055] Wherein, the first environmental feature is used to characterize the environmental features in the multiple image data at the first time moment, and the second environmental feature is used to characterize the environmental features in the multiple image data at the next time moment.

[0056] After the pedestrian tracking system acquires multiple image data of the first moment and multiple image data of the next moment, it will also acquire the first environmental feature in the multiple image data of the first moment and the second environmental feature in the multiple image data of the next moment. The first environmental feature is used to characterize the environmental features in the multiple image data of the first moment, and the second environmental feature is used to characterize the environmental features in the multiple image data of the next moment. The environmental features can be lighting features, brightness features, and background features (such as tree positions, leaf states, building positions, etc.).

[0057] Specifically, in dynamic scenes, pedestrian tracking operations face challenges from complex factors such as changing lighting, background interference, rapid movement, and occlusion. Therefore, this application's embodiments optimize the multimodal feature fusion and tracking process through real-time environment assessment and adaptive adjustment algorithms, enabling the system to maintain high accuracy and robustness in dynamic and changing environments. Adapting to dynamic scenes first requires real-time environment assessment to dynamically adjust various parts of the system (e.g., the weights of modal fusion, the parameters of the tracking algorithm, etc.). Environment assessment is mainly based on image content (such as lighting, movement speed, occlusion, etc.) and sensor inputs (such as detection code information, IMSI signal strength, etc.) to determine the degree of change in the current scene.

[0058] Step C2: Determine environmental change information based on the first environmental feature and the second environmental feature.

[0059] The environmental change information includes one or more of the following: visible light change information, infrared change information, and background change information.

[0060] After determining the first environmental features of the first image at the first moment and the second environmental features at the next moment, the pedestrian tracking system determines environmental change information based on the first and second environmental features. This environmental change information includes one or more of the following: visible light change information, infrared change information, and background change information. Background change information can include information such as changes in the position of leaves. In other words, the pedestrian tracking system determines environmental change information (such as changes in illumination, infrared changes, and background changes) based on the environmental features of the first image at the first moment and the environmental features at the next moment.

[0061] Step C3: Determine the preset weight coefficients and model parameters of the pre-trained state prediction model based on the environmental change information.

[0062] The preset weight coefficients and model parameters of the pre-trained state prediction model are different for different environmental change information.

[0063] After determining the background change information, the pedestrian tracking system determines the preset weight coefficients and pre-trained state prediction model parameters based on the environmental change information. The preset weight coefficients and pre-trained state prediction model parameters differ for different environmental changes. In other words, the pedestrian tracking system pre-stores the weight coefficients and state prediction model parameters corresponding to various environmental changes. This allows for dynamic adjustment of the weight coefficients and state prediction model parameters as the environment in the first frame continuously changes, ensuring that the weight coefficients and state prediction model parameters always match the newly changed environment, thus improving the accuracy of predicting user movement.

[0064] Specifically, based on environmental assessment results, we can dynamically adjust the weights of infrared images, visible light images, and detection information to enhance the system's adaptability to specific environments. Especially in environments with changing lighting, occlusion, and rapid movement, the weights of different modalities need to be adjusted in real time. Furthermore, in dynamic environments, pedestrian movement patterns may change; therefore, the pedestrian state information prediction method also needs to be adaptively optimized according to the current environment. To improve prediction accuracy, the pedestrian tracking system updates the model parameters of the state prediction model based on changes in environmental information.

[0065] Real-time feedback is not only an assessment of the current tracking status but also includes long-term tracking and optimization of system performance. By continuously collecting real-time feedback, the system can continuously improve its algorithm through online learning or incremental optimization, adapting to different scene and target changes. The pedestrian tracking system determines the tracking error based on the predicted and actual results and adjusts the model parameters according to the error changes through incremental learning algorithms. That is, it adjusts the model parameters based on real-time feedback to adapt to the continuous changes of pedestrians. Incremental learning enables the pedestrian tracking system to gradually optimize parameters over long-term tracking tasks, improving the system's adaptability to targets. Especially when the appearance or behavior of the target changes, the system can quickly adjust the model to maintain tracking accuracy.

[0066] Figure 2 This is a flowchart illustrating a second method for obtaining pedestrian traces provided in an embodiment of this specification, as shown below. Figure 2 As shown, the schematic diagram includes: Step 202: Acquire multiple image data of the first frame at the first moment, and fuse them based on preset weight coefficients to obtain the first fused image.

[0067] The first image is the image from which pedestrian traces are to be acquired, and the image data includes at least two of the following: visible light image, infrared image, and detection image.

[0068] Step 204: Determine the first state information of the first pedestrian in the first fused image.

[0069] Wherein, the first pedestrian is the pedestrian whose trace is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image.

[0070] Step 206: Based on the first state information, predict the state information of the first pedestrian at the next moment using a pre-trained state prediction model to obtain the predicted state information.

[0071] Step 208: Obtain multiple image data of the first frame at the next moment, and fuse them based on the preset weight coefficient to obtain a second fused image.

[0072] Step 210: Determine the second state information of the first pedestrian in the second fused image.

[0073] Step 212: Update the parameters of the pre-trained state prediction model based on the predicted state information and the second state information.

[0074] Step 214: Determine the pedestrian trail of the first pedestrian based on the second state information and the predicted state information.

[0075] The second state information refers to the state information of the first pedestrian in the second fused image.

[0076] Step 216: Obtain the first environmental features of the first screen at the first moment and the second environmental features at the next moment.

[0077] Wherein, the first environmental feature is used to characterize the environmental features in the multiple image data at the first time moment, and the second environmental feature is used to characterize the environmental features in the multiple image data at the next time moment.

[0078] Step 218: Determine environmental change information based on the first environmental feature and the second environmental feature.

[0079] The environmental change information includes one or more of the following: visible light change information, infrared change information, and background change information.

[0080] Step 220: Determine the preset weight coefficients and model parameters of the pre-trained state prediction model based on the environmental change information.

[0081] The preset weight coefficients and model parameters of the pre-trained state prediction model are different for different environmental change information.

[0082] In the embodiments described in the specification, the state information of the first pedestrian at the next moment is predicted by a pre-trained state prediction model, and the parameters of the state prediction model are updated based on the prediction results and the actual results. The model and weight coefficients are also updated by environmental change information, which can improve the adaptability of the pedestrian tracking system and increase the accuracy of pedestrian tracking.

[0083] In one implementation, steps D1-D2 may also be performed before determining the second state information of the first pedestrian in the second fused image (step 108): Step D1: Obtain the first appearance features of the first pedestrian and the second appearance features of the second pedestrian.

[0084] Wherein, the first appearance feature is used to characterize the appearance features of the first pedestrian, and the second pedestrian is any pedestrian in the second fused image, and the second appearance feature is used to characterize the appearance features of the second pedestrian.

[0085] The pedestrian tracking system acquires the first appearance features of the first pedestrian in the first frame at the first moment and the second appearance features of the second pedestrian in the next moment. The first appearance features are the appearance features of the first pedestrian, such as the first pedestrian's clothing and hairstyle. The second pedestrian is any pedestrian included in the first frame at the second moment. The second appearance features are used to characterize the appearance features of the second pedestrian, such as clothing and hairstyle.

[0086] Furthermore, the pedestrian tracking system can acquire the appearance features of the first pedestrian in the first frame at the first moment, and the feature information of all pedestrians included in the first frame at the next moment, that is, acquire the first appearance features of the first pedestrian and the second appearance features of multiple second pedestrians.

[0087] Step D2: The second appearance feature that has a similarity greater than the first appearance feature than the first appearance feature is identified as the third appearance feature, and the second pedestrian corresponding to the third appearance feature is identified as the first pedestrian.

[0088] After determining the first appearance feature of the first pedestrian and the second appearance feature of the second pedestrian, the pedestrian tracking acquisition system determines the similarity between the first appearance feature and the second appearance feature, and judges whether the similarity between the first appearance feature and the second appearance feature is greater than a first threshold. If so, the second appearance feature that is greater than the first threshold is determined as the third appearance feature, and finally the second pedestrian corresponding to the third appearance feature is determined as the first pedestrian.

[0089] Furthermore, the pedestrian tracking system can determine the appearance features of the first pedestrian at the first moment and the second appearance features of all pedestrians at the next moment. Then, it can determine whether there is a second appearance feature with a similarity greater than a first threshold to the first appearance feature. If so, the second pedestrian corresponding to the second appearance feature with a similarity greater than the first threshold is identified as the first pedestrian.

[0090] Specifically, when performing real-time tracking of the first pedestrian, the pedestrian tracking system first needs to identify the first pedestrian from multiple image data (or fused images) corresponding to each moment in the first frame in order to continuously acquire the first pedestrian's track. The pedestrian tracking system can determine the first pedestrian in the images corresponding to each moment through a pedestrian matching task. The pedestrian matching task calculates the similarity between the first pedestrian in the next moment and the first moment, thereby determining which target in the next moment corresponds to the target in the first moment based on the similarity. During the matching operation, the feature vectors of the fused image in the next moment and the fused image in the first moment are first determined. Then, features extracted by a deep convolutional neural network (CNN) are matched. Next, a target re-identification (ReID) model is used to extract the appearance features of the target through deep learning, and the similarity between the target in the next moment and the target in the first moment is calculated. In other words, the pedestrian tracking system can determine the first pedestrian by using the feature vectors and appearance features of each second pedestrian in the fused image in the next moment to determine their similarity to the first pedestrian, thus identifying the first pedestrian.

[0091] In one implementation, after obtaining the first appearance features of the first pedestrian and the second appearance features of the second pedestrian (step D1), step E may also be performed: Step E: When the similarity between the second appearance feature and the first appearance feature is less than a second threshold, the second fused image is supplemented based on the predicted state information to obtain the supplemented second fused image.

[0092] After acquiring the first appearance feature and the second appearance feature, the pedestrian tracking acquisition system determines the similarity between the first appearance feature and the second appearance feature, and whether the similarity between the first appearance feature and the second appearance feature is less than a second threshold. If so, the second fused image is supplemented based on the predicted state information to obtain the supplemented second fused image.

[0093] Furthermore, the pedestrian tracking system can determine the appearance features of the first pedestrian at the first moment and the second appearance features of all pedestrians at the next moment. It then determines whether the similarity between the second and first appearance features of all pedestrians is less than a second threshold. If so, it supplements the second fused image based on predicted state information, thus obtaining a supplemented second fused image. Specifically, when supplementing the second fused image based on predicted state information, the pedestrian tracking system can directly add the predicted state information to the second fused image, such as directly adding the predicted position and motion state information of the first pedestrian to the corresponding positions in the second fused image, thereby obtaining the supplemented second fused image.

[0094] Specifically, when the similarity between the physical characteristics of the second pedestrian and the first pedestrian is less than a second threshold, it indicates that the first pedestrian is missing or has been incorrectly tracked. In this case, the pedestrian tracking system can automatically trigger an alarm system through a real-time feedback mechanism to notify the operator to intervene. Simultaneously, the system can dynamically adjust the alarm threshold based on real-time assessment of tracking accuracy, ensuring that an alarm is only triggered in critical tracking failures.

[0095] In the real-time feedback and optimization phase, various algorithm parameters (such as modal fusion weights and Kalman filter parameters) are adjusted in real time by evaluating the current tracking results, enabling the system to adapt to changes in the dynamic environment and target behavior. The real-time feedback mechanism allows the system to continuously improve tracking accuracy through incremental learning and adaptive optimization, especially in complex scenarios such as changing lighting, background interference, and occlusion. Simultaneously, decision support and alarm mechanisms ensure timely alerts when tracking failures occur, guaranteeing stable system operation and efficient tracking.

[0096] Figure 3 This is a flowchart illustrating a third pedestrian tracking system provided in one embodiment of this specification, as shown below. Figure 3 As shown, the schematic diagram includes: Step 302: Acquire multiple image data of the first frame at the first moment, and fuse them based on preset weight coefficients to obtain the first fused image.

[0097] The first image is the image from which pedestrian traces are to be acquired, and the image data includes at least two of the following: visible light image, infrared image, and detection image.

[0098] Step 304: Determine the first state information of the first pedestrian in the first fused image, and predict the state information of the next moment based on the first state information to obtain the predicted state information.

[0099] Wherein, the first pedestrian is the pedestrian whose trace is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image.

[0100] Step 306: Obtain multiple image data of the first frame at the next moment, and fuse them based on the preset weight coefficient to obtain a second fused image.

[0101] Step 308: Obtain the first appearance features of the first pedestrian and the second appearance features of the second pedestrian.

[0102] Wherein, the first appearance feature is used to characterize the appearance features of the first pedestrian, and the second pedestrian is any pedestrian in the second fused image, and the second appearance feature is used to characterize the appearance features of the second pedestrian.

[0103] Step 310: The second appearance feature that has a similarity greater than the first appearance feature to the first appearance feature is identified as the third appearance feature, and the second pedestrian corresponding to the third appearance feature is identified as the first pedestrian.

[0104] Step 312: When the similarity between the second appearance feature and the first appearance feature is less than the second threshold, the second fused image is supplemented based on the predicted state information to obtain the supplemented second fused image.

[0105] Step 314: Determine the second state information of the first pedestrian in the second fused image, and determine the pedestrian trace of the first pedestrian based on the second state information and the predicted state information.

[0106] The second state information refers to the state information of the first pedestrian in the second fused image.

[0107] In the embodiment of the specification, by determining the similarity between the appearance features of the second pedestrian and the appearance features of the first pedestrian, the second pedestrian with a similarity greater than a first threshold is identified as the first pedestrian, and the first pedestrian's trace is acquired. When the similarity is less than a second threshold, the second fused image is supplemented, so that not only can the trace information of the first pedestrian be accurately acquired, but also an alarm can be triggered when the first pedestrian is lost or obscured.

[0108] It should be noted that the pedestrian tracking method provided in this application embodiment can be executed by a pedestrian tracking device or a control module within that device for executing the pedestrian tracking method. This application embodiment uses the execution of the pedestrian tracking method by a pedestrian tracking device as an example to illustrate the pedestrian tracking device provided in this application embodiment.

[0109] Figure 4 This is a schematic diagram of the pedestrian tracking device according to an embodiment of the present invention. Figure 4 As shown, the pedestrian tracking device includes: a first acquisition module 402, a first determination module 404, a second acquisition module 406, and a second determination module 408.

[0110] The first acquisition module 402 is used to acquire multiple image data of the first screen at a first moment, and fuse them based on a preset weight coefficient to obtain a first fused image. The first screen is the screen to be used for pedestrian trace acquisition. The image data includes at least two of visible light images, infrared images and detection images. The first determining module 404 is used to determine the first state information of the first pedestrian in the first fused image, and predict the state information of the next moment based on the first state information to obtain the predicted state information. The first pedestrian is the pedestrian in the first fused image whose pedestrian trace is to be acquired, and the first state information is the state information of the first pedestrian in the fused image. The second acquisition module 406 is used to acquire multiple image data of the first screen at the next moment, and perform feature fusion on them based on the preset weight coefficient to obtain a second fused image; The second determining module 408 is used to determine the second state information of the first pedestrian in the second fused image, and to determine the pedestrian trace of the first pedestrian based on the second state information and the predicted state information, wherein the second state information is the state information of the first pedestrian in the second fused image.

[0111] The pedestrian tracking device in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices can be servers, network-attached storage (NAS), personal computers (PCs), televisions (TVs), ATMs, or self-service machines, etc. This application embodiment does not impose specific limitations.

[0112] The pedestrian tracking device in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0113] The pedestrian tracking device provided in this application embodiment can achieve... Figures 1 to 3 The various processes implemented in the method embodiments are not described in detail here to avoid repetition.

[0114] Based on the same technical concept, embodiments of this application also provide an electronic device for performing the above-described pedestrian tracking method. Figure 5This is a schematic diagram of the structure of an electronic device to implement various embodiments of this application. The electronic device can vary significantly due to differences in configuration or performance, and may include a processor 502, a communications interface 504, a memory 506, and a communication bus 508. The processor 502, communications interface 504, and memory 506 communicate with each other via the communication bus 508. The processor 502 can call a computer program stored in the memory 506 and executable on the processor 502 to perform the following steps: Multiple image data of the first screen at a first moment are acquired, and feature fusion is performed on them based on a preset weight coefficient to obtain a first fused image. The first screen is the screen to be used for pedestrian tracking acquisition, and the image data includes at least two of visible light images, infrared images, and detection images. The first state information of the first pedestrian in the first fused image is determined, and the state information of the next moment is predicted based on the first state information to obtain the predicted state information. The first pedestrian is the pedestrian whose pedestrian trace is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image. Acquire multiple image data of the first frame at the next moment, and fuse them based on the preset weight coefficient to obtain a second fused image; The second state information of the first pedestrian in the second fused image is determined, and the pedestrian trace of the first pedestrian is determined based on the second state information and the predicted state information. The second state information is the state information of the first pedestrian in the second fused image.

[0115] In one implementation, acquiring multiple image data of the first frame at a first moment and fusing them based on preset weight coefficients to obtain a first fused image includes: The first environmental feature of the first frame at the first moment is obtained, and the first environmental feature is used to characterize the environmental features in multiple image data at the first moment. Based on a preset weight coefficient correspondence table, a first weight coefficient corresponding to the first environmental feature is determined, so as to perform feature fusion on multiple image data at the first time based on the first weight coefficient. The weight coefficient correspondence table stores the weight coefficients corresponding to different environmental features in advance.

[0116] In one implementation, the step of predicting the state information for the next time step based on the first state information to obtain predicted state information includes: Based on the first state information, the state information of the first pedestrian at the next moment is predicted by a pre-trained state prediction model to obtain the predicted state information. After determining the second state information of the first pedestrian in the second fused image, the method further includes: The parameters of the pre-trained state prediction model are updated based on the predicted state information and the second state information.

[0117] In one implementation, the method further includes: The first environmental feature of the first frame at the first moment and the second environmental feature at the next moment are obtained. The first environmental feature is used to characterize the environmental features in the multiple image data at the first moment, and the second environmental feature is used to characterize the environmental features in the multiple image data at the next moment. Based on the first environmental feature and the second environmental feature, environmental change information is determined, wherein the environmental change information includes one or more of visible light change information, infrared change information and background change information; The preset weight coefficients and model parameters of the pre-trained state prediction model are determined based on the environmental change information. The preset weight coefficients and model parameters of the pre-trained state prediction model are different for different environmental change information.

[0118] In one implementation, before determining the second state information of the first pedestrian in the second fused image, the method further includes: A first appearance feature of the first pedestrian and a second appearance feature of the second pedestrian are obtained. The first appearance feature is used to characterize the appearance feature of the first pedestrian. The second pedestrian is any pedestrian in the second fused image. The second appearance feature is used to characterize the appearance feature of the second pedestrian. The second appearance feature that has a similarity to the first appearance feature greater than a first threshold is identified as the third appearance feature, and the second pedestrian corresponding to the third appearance feature is identified as the first pedestrian.

[0119] In one implementation, after obtaining the first appearance feature of the first pedestrian and the second appearance feature of the second pedestrian, the method further includes: When the similarity between the second appearance feature and the first appearance feature is less than a second threshold, the second fused image is supplemented based on the predicted state information to obtain the supplemented second fused image.

[0120] The specific execution steps can be found in the various steps of the above-described pedestrian tracking method embodiment, and can achieve the same technical effect. To avoid repetition, they will not be repeated here.

[0121] It should be noted that the electronic devices in the embodiments of this application include: servers, terminals, or other devices besides terminals.

[0122] The above electronic device structure does not constitute a limitation on the electronic device. An electronic device may include more or fewer components than illustrated, or combine certain components, or arrange them differently. For example, an input unit may include a Graphics Processing Unit (GPU) and a microphone, and a display unit may use a liquid crystal display (LCD), organic light-emitting diode (OLED), or other similar display panels. User input units include at least one of a touch panel and other input devices. A touch panel is also called a touchscreen. Other input devices may include, but are not limited to, physical keyboards, function keys (such as volume control buttons, power buttons, etc.), trackballs, mice, and joysticks, which will not be elaborated further here.

[0123] Memory can be used to store software programs and various data. Memory can primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area can store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, memory can include volatile memory or non-volatile memory, or both. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0124] The processor may include one or more processing units; optionally, the processor integrates an application processor and a modem processor, wherein the application processor mainly handles operations related to the operating system, user interface, and applications, while the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into the processor.

[0125] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described pedestrian tracking method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0126] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0127] This application also provides a computer program product, which, when executed by a processor, implements the various processes of the above-described pedestrian tracking method embodiment and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0128] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described pedestrian tracking method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0129] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0130] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0131] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0132] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for obtaining pedestrian tracks, characterized in that, include: Multiple image data of the first screen at a first moment are acquired, and feature fusion is performed on them based on a preset weight coefficient to obtain a first fused image. The first screen is the screen to be used for pedestrian tracking acquisition, and the image data includes at least two of visible light images, infrared images, and detection images. The first state information of the first pedestrian in the first fused image is determined, and the state information of the next moment is predicted based on the first state information to obtain the predicted state information. The first pedestrian is the pedestrian whose pedestrian trace is to be acquired in the first fused image, and the first state information is the state information of the first pedestrian in the fused image. Acquire multiple image data of the first frame at the next moment, and fuse them based on the preset weight coefficient to obtain a second fused image; The second state information of the first pedestrian in the second fused image is determined, and the pedestrian trace of the first pedestrian is determined based on the second state information and the predicted state information. The second state information is the state information of the first pedestrian in the second fused image.

2. The method according to claim 1, characterized in that, The step of acquiring multiple image data of the first frame at a first moment and fusing them based on preset weight coefficients to obtain a first fused image includes: The first environmental feature of the first frame at the first moment is obtained, and the first environmental feature is used to characterize the environmental features in multiple image data at the first moment. Based on a preset weight coefficient correspondence table, a first weight coefficient corresponding to the first environmental feature is determined, so as to perform feature fusion on multiple image data at the first time based on the first weight coefficient. The weight coefficient correspondence table stores the weight coefficients corresponding to different environmental features in advance.

3. The method according to claim 1, characterized in that, The step of predicting the state information for the next time step based on the first state information to obtain predicted state information includes: Based on the first state information, the state information of the first pedestrian at the next moment is predicted by a pre-trained state prediction model to obtain the predicted state information. After determining the second state information of the first pedestrian in the second fused image, the method further includes: The parameters of the pre-trained state prediction model are updated based on the predicted state information and the second state information.

4. The method according to claim 3, characterized in that, The method further includes: The first environmental feature of the first frame at the first moment and the second environmental feature at the next moment are obtained. The first environmental feature is used to characterize the environmental features in the multiple image data at the first moment, and the second environmental feature is used to characterize the environmental features in the multiple image data at the next moment. Based on the first environmental feature and the second environmental feature, environmental change information is determined, wherein the environmental change information includes one or more of visible light change information, infrared change information and background change information; The preset weight coefficients and model parameters of the pre-trained state prediction model are determined based on the environmental change information. The preset weight coefficients and model parameters of the pre-trained state prediction model are different for different environmental change information.

5. The method according to claim 1, characterized in that, Before determining the second state information of the first pedestrian in the second fused image, the method further includes: A first appearance feature of the first pedestrian and a second appearance feature of the second pedestrian are obtained. The first appearance feature is used to characterize the appearance feature of the first pedestrian. The second pedestrian is any pedestrian in the second fused image. The second appearance feature is used to characterize the appearance feature of the second pedestrian. The second appearance feature that has a similarity to the first appearance feature greater than a first threshold is identified as the third appearance feature, and the second pedestrian corresponding to the third appearance feature is identified as the first pedestrian.

6. The method according to claim 5, characterized in that, After obtaining the first appearance features of the first pedestrian and the second appearance features of the second pedestrian, the method further includes: When the similarity between the second appearance feature and the first appearance feature is less than a second threshold, the second fused image is supplemented based on the predicted state information to obtain the supplemented second fused image.

7. A pedestrian tracking device, characterized in that, include: The first acquisition module is used to acquire multiple image data of the first screen at a first moment, and fuse them based on preset weight coefficients to obtain a first fused image. The first screen is the screen to be used for pedestrian tracking acquisition. The image data includes at least two of visible light images, infrared images and detection images. The first determining module is used to determine the first state information of the first pedestrian in the first fused image, and predict the state information of the next moment based on the first state information to obtain the predicted state information. The first pedestrian is the pedestrian in the first fused image whose pedestrian trace is to be acquired, and the first state information is the state information of the first pedestrian in the fused image. The second acquisition module is used to acquire multiple image data of the first screen at the next moment, and perform feature fusion on them based on the preset weight coefficient to obtain a second fused image; The second determining module is used to determine the second state information of the first pedestrian in the second fused image, and to determine the pedestrian trace of the first pedestrian based on the second state information and the predicted state information, wherein the second state information is the state information of the first pedestrian in the second fused image.

8. A computer device, characterized in that, The device includes: Processor; and A memory configured to store computer-executable instructions configured to be executed by the processor, the executable instructions including steps for performing the pedestrian tracking method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is used to store computer-executable instructions that cause the computer to perform the pedestrian tracking acquisition method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the pedestrian tracking method according to any one of claims 1 to 6.

Images