Intelligent target tracking data analysis system and method based on DSP

Abstract

The present application discloses a DSP-based intelligent target tracking data analysis system and method, which belongs to the field of target tracking technology. The present application collects image data, extracts visual features of the image to generate visual feature vectors, extracts motion features of the image to generate motion feature vectors, identifies interference objects and generates visual feature vectors and motion feature vectors of the interference objects, constructs a target similarity interference model, imports the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object, constructs an environmental interference model, imports the collected image environment data into the environmental interference model to evaluate the degree of environmental interference, constructs a target tracking effect evaluation model, imports target similarity interference and environmental interference into the target tracking effect evaluation model, evaluates the target tracking effect in combination with the continuity and stability of target tracking, and improves the target tracking effect evaluation capability.

Description

Technical Field

[0001] The present application relates to the field of target tracking technology, and more specifically to a DSP-based intelligent target tracking data analysis system and method. Background Art

[0002] With the rapid development of computer vision technology, target tracking has become a core technology in systems such as automated surveillance, intelligent transportation, and autonomous driving. Existing target tracking methods mostly rely on single visual features. This makes it difficult to ensure stable and high-precision tracking when dealing with complex interference in dynamic environments, and can easily lead to tracking drift or target loss. Most systems focus solely on tracking success, lacking quantitative analysis of factors such as continuity, stability, and interference effects during the tracking process. Existing technologies also fail to quantitatively model environmental changes and the characteristic strength of interfering objects, resulting in tracking evaluations that fail to accurately reflect actual performance in interference environments.

[0003] This application provides a DSP-based intelligent target tracking data analysis system and method, which effectively analyzes the target tracking effect in a complex interference environment, comprehensively considers the similarity of the visual and motion characteristics of the interfering object, as well as environmental factors such as illumination mutation, occlusion, and scene complexity, and constructs a tracking performance evaluation mechanism through feature extraction, interference modeling, and tracking stability analysis. This not only improves the system's anti-interference ability and tracking stability in changing scenarios, but also provides data support for subsequent tracking optimization and strategy adjustment, and can work stably in complex practical application scenarios. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, this application proposes a DSP-based intelligent target tracking data analysis system and method.

[0005] To achieve the above objectives, this application provides the following technical solutions:

[0006] The DSP-based intelligent target tracking data analysis system and method includes the following specific steps:

[0007] Collect image data, extract visual features of the image and generate visual feature vectors;

[0008] Extract motion features from the image to generate a motion feature vector, identify the interference object and generate a visual feature vector and a motion feature vector of the interference object;

[0009] Construct a target similarity interference model, import the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object;

[0010] Construct an environmental interference model and import the collected image environment data into the environmental interference model to evaluate the degree of environmental interference;

[0011] A target tracking effect evaluation model is constructed, target similarity interference and environmental interference are introduced into the target tracking effect evaluation model, and the target tracking effect is evaluated in combination with the continuity and stability of target tracking.

[0012] Preferably, the collecting of image data, extracting visual features from the image and generating a visual feature vector comprises the following specific steps:

[0013] S11. Collect image data through high-definition cameras and other sensors, pre-process the collected raw image data, including image denoising, normalization, enhancement, and other operations, and perform time synchronization and spatial registration on multi-sensor data to ensure the consistency of data from different sensors in time and space;

[0014] S12. Convert the image to the HSV color space and construct a normalized color histogram for the target area. Each channel is divided into a fixed number of intervals. The number of pixels in each interval is counted and normalized to form a color feature. Each color partition corresponds to one dimension.

[0015] S13. Grayscale the image, analyze the grayscale symbiosis relationship of the image in multiple directions based on the gray-level symbiosis matrix (GLCM), and extract four texture features of the image: contrast, correlation, energy, and entropy;

[0016] S14. Extract edge information of the image using the Canny algorithm, calculate the geometric moment and central moment of the image based on the edge area, obtain geometric invariant moments with rotation, translation, and scale invariance, construct edge shape features, and identify and calculate the center of the target edge;

[0017] S15. Normalize the color features, texture features, and edge shape features to generate a visual feature vector.

[0018] Preferably, the steps of extracting motion features from an image to generate a motion feature vector, identifying an interference object, and generating a visual feature vector and a motion feature vector of the interference object include the following specific steps:

[0019] S21. Obtain the position of the target tracking object in the current frame image and the previous frame image, perform grayscale processing on the target area, the velocity vector is the ratio of the difference between the current frame position and the previous frame position and time, the change in the motion direction angle is obtained by the inverse tangent function and the target tracking object position, and the acceleration is calculated by the velocity change rate between the two frames. The components of the velocity vector on the x-axis and y-axis, the change in the motion direction angle and the acceleration are used to generate a motion feature vector, where the velocity vector calculation formula is: , the calculation formula for the change in the angle of motion direction is: , , the acceleration calculation formula is: ,in, is the target center coordinate of the current frame, is the target center coordinate of the previous frame, is the time interval between two frames;

[0020] S22. With the current tracking target position as the center, expand the detection window by a certain proportion, and perform image segmentation and connected domain analysis within this area. The segmentation method is threshold segmentation based on color clustering, and regional segmentation based on color distribution and connected domain analysis to extract potential object areas, and filter out areas that are too small or too different from the target size. For each interference candidate area, extract and generate the visual feature vector and motion feature vector of the interference object, and maintain the same feature space as the feature vector of the current tracking target.

[0021] Preferably, the constructing of the target similarity interference model, importing the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object comprises the following specific steps:

[0022] S31, substituting the visual feature vectors of the target tracking object and the interference object into the visual feature cosine similarity calculation formula to calculate the similarity between the visual features of the target tracking object and the interference object, wherein the visual feature cosine similarity calculation formula is: ,in, is the visual feature vector of the target tracking object, is the visual feature vector of the interference object, is the magnitude of the vector;

[0023] S32, substituting the motion feature vectors of the target tracking object and the interference object into the motion feature similarity calculation formula to calculate the similarity between the motion features of the target tracking object and the interference object, wherein the motion feature similarity calculation formula is: ,in, is the motion feature vector of the target object in the current frame, is the motion feature vector of the interference object in the current frame, is the Euclidean distance between the motion feature vectors of the target tracking object and the interference object, which is used to measure the difference in motion behavior. is the maximum value of the modulus length of the two;

[0024] S33. Substituting the visual feature cosine similarity and the motion feature similarity into the target similarity interference calculation formula to calculate the target similarity interference degree, wherein the target similarity interference calculation formula is: ,in, and is the weight.

[0025] Preferably, the step of constructing an environmental interference model and importing the collected image environment data into the environmental interference model to evaluate the degree of environmental interference includes the following specific steps:

[0026] S41. Substitute the average brightness value of the current frame and the previous frame into the light interference calculation formula to evaluate the light interference, wherein the light interference calculation formula is: ,in, is the average brightness value of the current frame, is the average brightness value of the previous frame;

[0027] S42: Substitute the target effective area of ​​the current frame and the previous frame into the partial occlusion interference calculation formula to evaluate the partial occlusion interference, wherein the partial occlusion interference calculation formula is: ,in, is the effective target area detected in the current frame, is the effective area of ​​the target in the previous frame, Standard area, the standard area is a constant, in order to avoid the denominator being 0;

[0028] S43. Substitute the number of edge pixels of the image into the scene complexity calculation formula to evaluate the scene complexity, wherein the scene complexity calculation formula is: ,in, is the total number of edge pixels detected in the image, extracted by the Canny edge detection algorithm, W is the image width, H is the image height, is the total number of pixels, is the number of significant moving or static objects in the scene, obtained by background subtraction statistics, Used to suppress order of magnitude differences and smooth complexity growth, Quantify the number of edges per unit area to reflect the complexity of the texture;

[0029] S44. Substitute illumination interference, partial occlusion interference, and scene complexity into the environmental interference calculation formula to evaluate the degree of environmental interference. The environmental interference calculation formula is: , where r is the interference factor, including illumination interference, partial occlusion interference and scene complexity, is the interference factor weight, is the theoretical minimum value of the rth interference factor, The theoretical maximum value of the rth interference factor, linear normalization eliminates dimensional differences.

[0030] Preferably, the target tracking effect evaluation model is constructed, target similarity interference and environmental interference are introduced into the target tracking effect evaluation model, and the target tracking effect is evaluated in combination with the continuity and stability of target tracking, including the following specific steps:

[0031] S51. Substitute the target similarity interference and the environmental interference into the comprehensive interference calculation formula to evaluate the comprehensive interference situation, wherein, in the comprehensive interference calculation formula: , k is the scaling factor used to control the steepness of the curve, Y is the total number of frames in the target tracking process, is the target similarity interference at frame t, The environmental interference in the t-th frame is generated by fusing the target similarity interference value and the environmental interference value, and mapping them through the Sigmoid function to generate the normalized comprehensive interference. The integral is used to accumulate the interference intensity of each frame in the entire tracking process;

[0032] S52: Substitute the actual center coordinates of the target tracking object and the center coordinates of the tracking frame into the target tracking stability calculation formula to evaluate the stability of the target tracking process. The target tracking stability calculation formula is: ,in, is the true target center coordinate, is the center coordinate of the tracking frame, is the diagonal length of the image;

[0033] S53, substituting the comprehensive interference, target tracking stability, and target tracking duration into the target tracking effect evaluation formula to evaluate the target tracking effect, wherein the target tracking effect evaluation formula is: ,where Z is the target tracking interruption duration, T is the total duration from the target appearance to disappearance, For the continuity of target tracking, the target tracking effect evaluation is compared with the threshold. If the target tracking effect value is higher than the threshold, the target tracking effect meets the requirements. If the target tracking effect value is lower than the threshold, it indicates that the target tracking effect does not meet the requirements.

[0034] The DSP-based intelligent target tracking data analysis system is implemented based on the above-mentioned DSP-based intelligent target tracking data analysis method, and specifically includes:

[0035] Image acquisition module, used for collecting image data and performing preprocessing;

[0036] A feature extraction module is used to extract visual features and motion features from an image and generate visual feature vectors and motion feature vectors;

[0037] A target similarity interference evaluation module is used to evaluate the similarity between the target tracking object and the interference object through the visual feature vector and motion feature vector of the target tracking object and the interference object;

[0038] An environmental interference assessment module is used to assess the degree of environmental interference through the collected image environmental data;

[0039] The target tracking effect evaluation module is used to evaluate the target tracking effect by combining target similarity interference and environmental interference with the continuity and stability of target tracking.

[0040] An electronic device comprises: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0041] The processor executes the above-mentioned DSP-based intelligent target tracking data analysis method by calling the computer program stored in the memory.

[0042] A computer-readable storage medium is characterized in that it stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned DSP-based intelligent target tracking data analysis method.

[0043] Compared with the prior art, the present invention has the following advantages:

[0044] The present application collects image data, extracts visual features of the image and generates visual feature vectors, extracts motion features of the image and generates motion feature vectors, identifies interference objects and generates visual feature vectors and motion feature vectors of the interference objects, constructs a target similarity interference model, imports the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object, constructs an environmental interference model, imports the collected image environment data into the environmental interference model to evaluate the degree of environmental interference, constructs a target tracking effect evaluation model, imports target similarity interference and environmental interference into the target tracking effect evaluation model, evaluates the target tracking effect in combination with the continuity and stability of target tracking, and improves the evaluation capability of the target tracking effect in an interference environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the overall process of the DSP-based intelligent target tracking data analysis method of this application;

[0046] Figure 2 Similar interference calculation flow chart for this application target;

[0047] Figure 3 A flow chart for tracking the effectiveness evaluation of this application goal;

[0048] Figure 4 This is a schematic diagram of the overall framework of the DSP-based intelligent target tracking data analysis system in this application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0050] Example 1

[0051] See also Figure 1-3 , this application provides an embodiment: a DSP-based intelligent target tracking data analysis method, which includes the following specific steps:

[0052] Collect image data, extract visual features of the image and generate visual feature vectors;

[0053] Extract motion features from the image to generate a motion feature vector, identify the interference object and generate a visual feature vector and a motion feature vector of the interference object;

[0054] Construct a target similarity interference model, import the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object;

[0055] Construct an environmental interference model and import the collected image environment data into the environmental interference model to evaluate the degree of environmental interference;

[0056] A target tracking effect evaluation model is constructed, target similarity interference and environmental interference are introduced into the target tracking effect evaluation model, and the target tracking effect is evaluated in combination with the continuity and stability of target tracking.

[0057] In this embodiment, it should be specifically explained that collecting image data, extracting visual features from the image, and generating a visual feature vector include the following specific steps:

[0058] S11. Collect image data through high-definition cameras and other sensors, pre-process the collected raw image data, including image denoising, normalization, enhancement, and other operations, and perform time synchronization and spatial registration on multi-sensor data to ensure the consistency of data from different sensors in time and space;

[0059] For example, for image data collected by the camera, a Gaussian filter algorithm is used to remove noise in the image and improve image clarity. For radar and sensor data, coordinate transformation and time calibration are performed to ensure accurate correspondence between data from different devices.

[0060] S12. Convert the image to the HSV color space, construct histograms for the three channels of hue, saturation, and brightness for the target area, segment each channel according to the set number of intervals, count the number of pixels in each segment, construct a frequency histogram, normalize the histogram of each channel, and the value of each interval represents its relative proportion in the target area. The normalized histograms of the three channels are sequentially spliced ​​to form a one-dimensional color feature vector, in which each dimension reflects the proportion of the color partition in the target area;

[0061] S13. Grayscale the image and analyze the grayscale co-occurrence relationship of the image in multiple directions based on the grayscale co-occurrence matrix (GLCM). Extract the four texture features of the image: contrast, correlation, energy, and entropy. Select a certain pixel distance (1 pixel) and direction (0 degrees, 45 degrees, 90 degrees, 135 degrees) to scan the entire grayscale image. Count the frequency of the grayscale value combination of any two pixels in the image at each direction and distance. For example, if we check the grayscale value of every two adjacent pixels in the 0-degree direction, we will obtain a matrix that represents the frequency of pixel pairs with different grayscale values ​​in the image. Based on the grayscale co-occurrence matrix, extract the four texture features of the image: contrast, correlation, energy, and entropy. Contrast measures the roughness of the image, correlation measures the relationship or consistency between grayscale values, energy measures the uniformity of the image texture, and entropy measures the complexity or uncertainty of the image grayscale distribution.

[0062] S14. Extract edge information from the image using the Canny algorithm, which includes noise removal, gradient calculation, non-maximum suppression, and double thresholding steps to obtain a clear edge map. Calculate the geometric moment and central moment of the image based on the edge region to obtain geometric invariant moments that are rotationally, translationally, and scale invariant. The geometric moment is a mathematical tool used to describe the shape of the target, reflecting the shape characteristics of the object in the image. The central moment is a normalized processing of the geometric moment to form the edge shape feature. Calculate the weighted coordinates of each pixel, where the weight value is the grayscale value of the pixel. Sum the weighted coordinates of all pixels to obtain the center of mass position of the target region, i.e., the target center.

[0063] S15. Normalize the color features, texture features, and edge shape features to generate a visual feature vector.

[0064] In this embodiment, it should be specifically explained that extracting motion features from an image to generate a motion feature vector, identifying an interference object, and generating a visual feature vector and a motion feature vector of the interference object include the following specific steps:

[0065] S21. Obtain the position of the target tracking object in the current frame image and the previous frame image, perform grayscale processing on the target area, the velocity vector is the ratio of the difference between the current frame position and the previous frame position and time, the change in the motion direction angle is obtained by the inverse tangent function and the target tracking object position, and the acceleration is calculated by the velocity change rate between the two frames. The components of the velocity vector on the x-axis and y-axis, the change in the motion direction angle and the acceleration are used to generate a motion feature vector, where the velocity vector calculation formula is: , the calculation formula for the change in the angle of motion direction is: , , the acceleration calculation formula is: ,in, is the target center coordinate of the current frame, is the target center coordinate of the previous frame, is the time interval between two frames;

[0066] S22. With the current tracking target position as the center, expand the detection window by a certain proportion, and perform image segmentation and connected domain analysis within this area. The segmentation method is threshold segmentation based on color clustering, and regional segmentation based on color distribution and connected domain analysis to extract potential object areas, and filter out areas that are too small or too different from the target size. For each interference candidate area, extract and generate the visual feature vector and motion feature vector of the interference object, and maintain the same feature space as the feature vector of the current tracking target.

[0067] Exemplarily, interference objects that affect tracking stability are detected in each frame of the image. A detection window of a set proportion (1.5 times the width and height) is expanded outward with the current tracking target position as the center, and the image in the window is preliminarily segmented based on color distribution. Several connected domain areas are extracted, and for each connected domain, screening is performed based on the area, aspect ratio and similarity with the target size. Interference objects with a scale close to the target are retained, and for each interference object, visual feature vectors and motion feature vectors are extracted according to predefined visual feature and motion feature extraction methods.

[0068] In this embodiment, it should be specifically explained that constructing a target similarity interference model and importing the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object includes the following specific steps:

[0069] S31, substituting the visual feature vectors of the target tracking object and the interference object into the visual feature cosine similarity calculation formula to calculate the similarity between the visual features of the target tracking object and the interference object, wherein the visual feature cosine similarity calculation formula is: ,in, is the visual feature vector of the target tracking object, is the visual feature vector of the interference object, is the magnitude of the vector;

[0070] S32, substituting the motion feature vectors of the target tracking object and the interference object into the motion feature similarity calculation formula to calculate the similarity between the motion features of the target tracking object and the interference object, wherein the motion feature similarity calculation formula is: ,in, is the motion feature vector of the target object in the current frame, is the motion feature vector of the interference object in the current frame, is the Euclidean distance between the motion feature vectors of the target tracking object and the interference object, which is used to measure the difference in motion behavior. The maximum value of the modulus of the two is normalized so that the distance value is related to the movement magnitude. The closer the similarity is to 1, the closer the movement behavior is and the stronger the interference is.

[0071] S33. Substituting the visual feature cosine similarity and the motion feature similarity into the target similarity interference calculation formula to calculate the target similarity interference degree, wherein the target similarity interference calculation formula is: ,in, and is the weight;

[0072] For example, in this embodiment, , indicating that the cosine similarity of visual features has a greater weight in the target similarity interference calculation and the interference effect is stronger.

[0073] In this embodiment, it should be specifically explained that constructing an environmental interference model and importing the collected image environment data into the environmental interference model to evaluate the degree of environmental interference includes the following specific steps:

[0074] S41. Substitute the average brightness value of the current frame and the previous frame into the light interference calculation formula to evaluate the light interference, wherein the light interference calculation formula is: ,in, is the average brightness value of the current frame, is the average brightness value of the previous frame;

[0075] S42: Substitute the target effective area of ​​the current frame and the previous frame into the partial occlusion interference calculation formula to evaluate the partial occlusion interference, wherein the partial occlusion interference calculation formula is: ,in, is the effective target area detected in the current frame, is the effective area of ​​the target in the previous frame, Standard area, the standard area is a constant, in order to avoid the denominator being 0;

[0076] S43. Substitute the number of edge pixels of the image into the scene complexity calculation formula to evaluate the scene complexity, wherein the scene complexity calculation formula is: ,in, is the total number of edge pixels detected in the image, extracted by the Canny edge detection algorithm, W is the image width, H is the image height, is the total number of pixels, and the impact of resolution differences is eliminated by normalizing the number of edges. is the number of significant moving or static objects in the scene, obtained by background subtraction statistics, Used to suppress order of magnitude differences and smooth complexity growth, Quantify the number of edges per unit area to reflect the complexity of the texture;

[0077] S44. Substitute illumination interference, partial occlusion interference, and scene complexity into the environmental interference calculation formula to evaluate the degree of environmental interference. The environmental interference calculation formula is: , where r is the interference factor, including illumination interference, partial occlusion interference and scene complexity, is the interference factor weight, is the theoretical minimum value of the rth interference factor, The theoretical maximum value of the rth interference factor, linear normalization eliminates dimensional differences;

[0078] For example, the illumination interference of the current frame is 60, the theoretical range is [30, 90], the partial occlusion interference is 0.4, the theoretical range is [0, 1], the scene complexity is 0.75, the theoretical range is [0.2, 1.0], and the interference factor weights are 0.4, 0.35 and 0.25 respectively. Substitute them into the environmental interference calculation formula to calculate the degree of environmental interference.

[0079] In this embodiment, it should be specifically explained that constructing a target tracking effect evaluation model, introducing target similarity interference and environmental interference into the target tracking effect evaluation model, and evaluating the target tracking effect in combination with the continuity and stability of target tracking include the following specific steps:

[0080] S51. Substitute the target similarity interference and the environmental interference into the comprehensive interference calculation formula to evaluate the comprehensive interference situation, wherein, in the comprehensive interference calculation formula: , k is the scaling factor used to control the steepness of the curve, Y is the total number of frames in the target tracking process, is the target similarity interference at frame t, For the environmental interference at the tth frame, the normalized comprehensive interference is generated by fusing the target similarity interference value and the environmental interference value and mapping them through the Sigmoid function. The integral is used to accumulate the interference intensity of each frame in the entire tracking process to comprehensively evaluate the comprehensive interference level that the system withstands in the full frame sequence;

[0081] Exemplarily, in this embodiment, the scaling factor is set to k=5, which increases the sensitivity to medium and high interference levels;

[0082] S52: Substitute the actual center coordinates of the target tracking object and the center coordinates of the tracking frame into the target tracking stability calculation formula to evaluate the stability of the target tracking process. The target tracking stability calculation formula is: ,in, is the true target center coordinate, is the center coordinate of the tracking frame, The length of the image diagonal is calculated by normalizing the tracking error of each frame, taking the average and the inverse, to measure the overall stability of the tracking process. The smaller the error, the higher the stability score.

[0083] S53, substituting the comprehensive interference, target tracking stability, and target tracking duration into the target tracking effect evaluation formula to evaluate the target tracking effect, wherein the target tracking effect evaluation formula is: ,where Z is the target tracking interruption duration, T is the total duration from the target appearance to disappearance, For the continuity of target tracking, the target tracking effect evaluation is compared with the threshold. If the target tracking effect value is higher than the threshold, the target tracking effect meets the requirements. If the target tracking effect value is lower than the threshold, it indicates that the target tracking effect does not meet the requirements.

[0084] For example, if the target tracking effect value is lower than the threshold and it is detected that the tracking performance has decreased due to increased environmental interference, the weights in the feature fusion strategy are dynamically adjusted according to the type of interference. When the lighting changes significantly, the weight of the color feature is reduced, and the weight of the texture and edge features is increased. When occlusion occurs, the importance of motion features in the matching process is enhanced, and the dependence on static visual features is weakened. When the continuity score decreases, indicating that the target is temporarily lost, the search area of ​​the next frame is adaptively expanded to increase the probability of recapture the target.

[0085] It should be noted here that the various setting parameters in this embodiment are obtained by obtaining representative target tracking interference data and target tracking effect conditions, hiring experts to manually judge whether the target tracking meets the requirements, and substituting the obtained historical data into the calculation results and judgment results of each step in this embodiment into the fitting software to output the values ​​of various setting parameters that meet the highest judgment accuracy.

[0086] The advantages of this embodiment over the prior art are as follows: the present application collects image data, extracts visual features of the image and generates a visual feature vector, extracts motion features of the image and generates a motion feature vector, identifies interference objects and generates visual feature vectors and motion feature vectors of the interference objects, constructs a target similarity interference model, imports the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the degree of similarity between the target tracking object and the interference object, constructs an environmental interference model, imports the collected image environment data into the environmental interference model to evaluate the degree of environmental interference, constructs a target tracking effect evaluation model, imports target similarity interference and environmental interference into the target tracking effect evaluation model, evaluates the target tracking effect in combination with the continuity and stability of target tracking, and improves the evaluation capability of the target tracking effect in an interference environment.

[0087] Example 2

[0088] like Figure 4 As shown, a DSP-based intelligent target tracking data analysis system is implemented based on the above-mentioned DSP-based intelligent target tracking data analysis method, which specifically includes an image acquisition module, a feature extraction module, a target similarity interference assessment module, an environmental interference assessment module and a target tracking effect assessment module. The image acquisition module is used to acquire image data and perform preprocessing; the feature extraction module is used to extract visual features and motion features of the image and generate visual feature vectors and motion feature vectors; the target similarity interference assessment module is used to assess the similarity between the target tracking object and the interference object through the visual feature vectors and motion feature vectors of the target tracking object and the interference object; the environmental interference assessment module is used to assess the degree of environmental interference through the collected image environment data; the target tracking effect assessment module is used to assess the target tracking effect through target similarity interference and environmental interference combined with the continuity and stability of target tracking.

[0089] It should be specifically noted here that the image acquisition module has a built-in image preprocessing unit to complete preliminary grayscale equalization, noise suppression, edge enhancement and other operations. The image acquisition module supports multi-resolution image acquisition mode, and the image data will be directly transferred to the DSP internal storage unit through the DMA channel or cache, reducing data transmission delay and improving the overall response speed of the system.

[0090] Example 3

[0091] This embodiment provides an electronic device, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;

[0092] The processor executes the above-mentioned DSP-based intelligent target tracking data analysis method by calling the computer program stored in the memory.

[0093] This electronic device can vary significantly depending on its configuration or performance, and can include one or more processors (Central Processing Units, CPUs) and one or more memories, wherein the memories store at least one computer program, which is loaded and executed by the processor to implement the DSP-based intelligent target tracking data analysis method provided in the above-mentioned method embodiment. The electronic device can also include other components for implementing the device's functions. For example, the electronic device can also include components such as wired or wireless network interfaces and input / output interfaces for data input and output. This embodiment is not described in detail here.

[0094] Example 4

[0095] This embodiment provides a computer-readable storage medium having a rewritable computer program stored thereon;

[0096] When the computer program runs on a computer device, the computer device is enabled to execute the above-mentioned DSP-based intelligent target tracking data analysis method.

[0097] The above embodiments can be implemented in whole or in part via software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. A computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions according to the embodiments of the present application are fully or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, computer instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via a wired network or / and a wireless network. A computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

Claims

1. The intelligent target tracking data analysis method based on DSP is characterized by: It includes the following specific steps: Collect image data, extract visual features of the image and generate visual feature vectors; Extract motion features from the image to generate a motion feature vector, identify the interference object and generate a visual feature vector and a motion feature vector of the interference object; Construct a target similarity interference model, import the visual feature vectors and motion feature vectors of the target tracking object and the interference object into the target similarity interference model to evaluate the similarity between the target tracking object and the interference object, including the following specific steps: substitute the visual feature vectors of the target tracking object and the interference object into the visual feature cosine similarity calculation formula to calculate the similarity between the visual features of the target tracking object and the interference object, wherein the visual feature cosine similarity calculation formula is: ,in, is the visual feature vector of the target tracking object, is the visual feature vector of the interference object, The motion feature vectors of the target tracking object and the interference object are substituted into the motion feature similarity calculation formula to calculate the similarity between the motion features of the target tracking object and the interference object. The motion feature similarity calculation formula is: ,in, is the motion feature vector of the target object in the current frame, is the motion feature vector of the interference object in the current frame, is the Euclidean distance between the motion feature vectors of the target tracking object and the interference object. The visual feature cosine similarity and motion feature similarity are substituted into the target similarity interference calculation formula to calculate the target similarity interference degree. The target similarity interference calculation formula is: ,in, and is the weight; Construct an environmental interference model and import the collected image environment data into the environmental interference model to evaluate the degree of environmental interference. The specific steps include: substituting the average brightness value of the current frame and the previous frame into the light interference calculation formula to evaluate the light interference. The light interference calculation formula is: ,in, is the average brightness value of the current frame, is the average brightness value of the previous frame, and the target effective area of ​​the current frame and the previous frame are substituted into the partial occlusion interference calculation formula to evaluate the partial occlusion interference. The partial occlusion interference calculation formula is: ,in, The effective target area detected in the current frame, is the effective area of ​​the target in the previous frame, As the standard area, the number of image edge pixels is substituted into the scene complexity calculation formula to evaluate the scene complexity, where the scene complexity calculation formula is: ,in, is the total number of edge pixels detected in the image, W is the image width, H is the image height, is the total number of pixels, is the number of significant moving and static objects in the scene. The degree of environmental interference is evaluated by substituting illumination interference, partial occlusion interference, and scene complexity into the environmental interference calculation formula. The environmental interference calculation formula is: , where r is the interference factor, including illumination interference, partial occlusion interference and scene complexity, is the interference factor weight, is the theoretical minimum value of the rth interference factor, The theoretical maximum value of the rth interference factor; A target tracking effect evaluation model is constructed, target similarity interference and environmental interference are introduced into the target tracking effect evaluation model, and the target tracking effect is evaluated in combination with the continuity and stability of target tracking.

2. The DSP-based intelligent target tracking data analysis method according to claim 1, characterized in that: The steps of collecting image data, extracting visual features from the image and generating a visual feature vector include the following specific steps: S11, collecting image data through high-definition cameras and other sensors, pre-processing the collected raw image data, and performing time synchronization and spatial registration on multi-sensor data; S12. Convert the image to the HSV color space and construct a normalized color histogram for the target area. Each channel is divided into a fixed number of intervals. The number of pixels in each interval is counted and normalized to form a color feature. Each color partition corresponds to one dimension. S13, graying the image, analyzing the gray-level symbiosis relationship of the image in multiple directions based on the gray-level symbiosis matrix, and extracting four texture features of the image: contrast, correlation, energy, and entropy; S14. Extract edge information of the image using the Canny algorithm, calculate the geometric moment and central moment of the image based on the edge area, obtain geometric invariant moments with rotation, translation, and scale invariance, construct edge shape features, and identify and calculate the center of the target edge; S15. Normalize the color features, texture features, and edge shape features to generate a visual feature vector.

3. The DSP-based intelligent target tracking data analysis method according to claim 2, characterized in that: The steps of extracting motion features from an image to generate a motion feature vector, identifying an interfering object, and generating a visual feature vector and a motion feature vector of the interfering object include the following specific steps: S21. Obtain the position of the target tracking object in the current frame image and the previous frame image, perform grayscale processing on the target area, the velocity vector is the ratio of the difference between the current frame position and the previous frame position and time, the change in the angle of motion direction is obtained by the inverse tangent function and the position of the target tracking object, and the acceleration is calculated by the rate of change of velocity between the two frames, and the components of the velocity vector on the x-axis and y-axis, the change in the angle of motion direction and the acceleration are used to generate a motion feature vector; S22. With the current tracking target position as the center, expand the detection window by a certain proportion, perform image segmentation and connected domain analysis in this area, extract potential interference object areas, and extract and generate visual feature vectors and motion feature vectors of the interference objects.

4. The DSP-based intelligent target tracking data analysis method according to claim 3, characterized in that: The target tracking effect evaluation model is constructed, target similarity interference and environmental interference are introduced into the target tracking effect evaluation model, and the target tracking effect is evaluated in combination with the continuity and stability of target tracking. The specific steps include: S51. Substitute the target similarity interference and the environmental interference into the comprehensive interference calculation formula to evaluate the comprehensive interference situation, wherein, in the comprehensive interference calculation formula: , k is the scaling factor, Y is the total number of frames in the target tracking process, is the target similarity interference at frame t, is the environmental interference at frame t; S52: Substitute the actual center coordinates of the target tracking object and the center coordinates of the tracking frame into the target tracking stability calculation formula to evaluate the stability of the target tracking process. The target tracking stability calculation formula is: ,in, is the true target center coordinate, is the center coordinate of the tracking frame, is the diagonal length of the image; S53, substituting the comprehensive interference, target tracking stability, and target tracking duration into the target tracking effect evaluation formula to evaluate the target tracking effect, wherein the target tracking effect evaluation formula is: ,where Z is the target tracking interruption duration, T is the total duration from the target appearance to disappearance, For the continuity of target tracking, the target tracking effect evaluation is compared with the threshold. If the target tracking effect value is higher than the threshold, the target tracking effect meets the requirements. If the target tracking effect value is lower than the threshold, it indicates that the target tracking effect does not meet the requirements.

5. A DSP-based intelligent target tracking data analysis system, which is implemented based on the DSP-based intelligent target tracking data analysis method according to any one of claims 1 to 4, characterized in that: Specifically include: Image acquisition module, used for collecting image data and performing preprocessing; A feature extraction module is used to extract visual features and motion features from an image and generate visual feature vectors and motion feature vectors; A target similarity interference evaluation module is used to evaluate the similarity between the target tracking object and the interference object through the visual feature vector and motion feature vector of the target tracking object and the interference object; An environmental interference assessment module is used to assess the degree of environmental interference through the collected image environmental data; The target tracking effect evaluation module is used to evaluate the target tracking effect by combining target similarity interference and environmental interference with the continuity and stability of target tracking.

6. An electronic device comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor; It is characterized in that the processor executes the DSP-based intelligent target tracking data analysis method as described in any one of claims 1 to 4 by calling the computer program stored in the memory.

7. A computer-readable storage medium, characterized in that Instructions are stored, and when the instructions are run on a computer, the computer is caused to execute the DSP-based intelligent target tracking data analysis method according to any one of claims 1 to 4.

Images