Visual sensing personnel positioning and tracking system and method

Abstract

The invention relates to the technical field of visual positioning, in particular to a visual sensing personnel positioning and tracking system and method, and adopts the technical scheme that the system comprises a visual acquisition module, a data preprocessing module, a target detection and feature extraction module and a positioning calculation and tracking module. The positioning calculation and tracking module establishes a self-adaptive motion model according to the motion mode of the target, can accurately predict and track the position change of the target, effectively deals with various motion conditions of personnel, ensures the continuity and accuracy of tracking, avoids the loss of the target, and improves the accuracy of tracking when the motion model is established. Interference factors such as noise points and spots in the image can be effectively removed through image processing, the accuracy of follow-up target detection and feature extraction is improved, a gray histogram of an original image is changed from a relatively concentrated gray interval to uniform distribution in a whole gray range, the dynamic range of the image is effectively expanded, and the accuracy of target detection and feature extraction is improved. And the global contrast of the image is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of visual positioning technology, and in particular to a visual sensing personnel positioning and tracking system and method. Background Art

[0002] In today's personnel positioning and tracking field, with the continuous growth of security needs and the widespread application of intelligent monitoring systems, it has become crucial to accurately and efficiently obtain the location and movement information of personnel in a specific area.

[0003] Traditional personnel positioning technology has many limitations. Early visual acquisition methods based on a single camera are prone to problems such as target occlusion and blind spots in complex monitoring scenarios due to their single viewing angle, making it impossible to fully and accurately capture the position and posture changes of personnel.

[0004] In addition, traditional image processing technology is also powerless when dealing with image data in complex environments. A large number of raw images that have not been optimized often contain a lot of noise. Traditional simple filtering methods are difficult to remove multiple types of noise at the same time, and the process of noise removal is prone to blurring of image edges and details, which is not conducive to subsequent target detection and feature extraction. Summary of the invention

[0005] The purpose of the present invention is to provide a visual sensing personnel positioning and tracking system and method to overcome the deficiencies of the prior art and achieve high-precision, stable and reliable positioning and tracking of personnel targets.

[0006] The technical solution of the present invention is a visual sensing personnel positioning and tracking system, comprising: The visual acquisition module consists of multiple binocular cameras with high resolution, high frame rate and different viewing angles distributed in the monitoring area. The binocular cameras can automatically adjust the acquisition parameters based on the ambient light intensity and transmit the collected image data to the data preprocessing module in real time through a high-speed data transmission interface; The data preprocessing module is used to perform grayscale conversion, filtering and denoising, and histogram equalization processing on the received image data; The target detection and feature extraction module detects human targets from the processed image data based on the target detection algorithm, and extracts the LBP texture features and the color features after PCA dimensionality reduction of each human target; The positioning calculation and tracking module uses binocular vision principles and stereo vision algorithms to calculate the position coordinates of personnel targets in three-dimensional space, and establishes an adaptive motion model based on the position coordinates.

[0007] Preferably, the binocular camera in the visual acquisition module has the function of automatically adjusting focus, aperture and white balance, and the data transmission interface is a Gigabit Ethernet interface or a fiber optic interface.

[0008] Preferably, the grayscale processing in the data preprocessing module is based on the weighted average method, specifically, the values ​​of the three RGB channels of each pixel in the image data collected by the binocular camera are multiplied by preset weight coefficients, the preset weight coefficient of the red channel is 0.299, the preset weight coefficient of the green channel is 0.587, and the preset weight coefficient of the blue channel is 0.114. The weighted values ​​of the three channels are added to obtain the grayscale value of the pixel to reduce the amount of data and highlight the contour features of the target person.

[0009] Preferably, the filtering and denoising steps in the data preprocessing module are as follows: S1: Perform median filtering on the grayscale image data. The median filter uses a window size of 3x3 pixels. Starting from the upper left corner of the image data, the image data is traversed pixel by pixel in units of windows. For each pixel point, all pixel grayscale values ​​in its neighborhood window are collected and sorted. The sorted middle value is used as the median filter grayscale value of the pixel point after median filtering to remove salt and pepper noise. S2: Gaussian filtering is performed on the image data after median filtering. The standard deviation of Gaussian filtering is set according to the noise situation of the image data and the required smoothness. The standard deviation is between 0.5-2, and the convolution kernel size is 6σ+1. The Gaussian kernel generated by the two-dimensional Gaussian distribution function is used to perform convolution operation on the image data. Specifically, the center of the Gaussian kernel is aligned with each pixel in the image data, and the weighted sum of the Gaussian kernel and the pixels in the neighborhood of the pixel is calculated. The weighted sum is used as the Gaussian filter grayscale value to remove Gaussian noise and retain the basic edge and contour information of the person target to avoid excessive blurring of the target edge.

[0010] Preferably, the histogram equalization processing in the data preprocessing module is specifically to perform histogram equalization processing on the image data after filtering and denoising, so as to make the gray value distribution of the image data more uniform and enhance the contrast of the image data; Specifically, the grayscale value range of the filtered and denoised image data is divided into 256 sub-intervals, and the number of pixels in each sub-interval is counted to obtain a grayscale value histogram. The cumulative distribution function of each gray level is calculated based on the grayscale value histogram. , for each gray level, The specific calculation method is: ; in, is the new gray value, is the cumulative distribution function value of the current gray level, is the minimum value in the cumulative distribution function, are the number of rows and columns of image data respectively.

[0011] Preferably, the positioning calculation and tracking module adopts binocular vision principle and stereo vision algorithm to calculate the position coordinates of the personnel target in three-dimensional space, and the camera in the visual acquisition module is a binocular camera, which specifically includes the following steps: S100: The binocular camera uses the Zhang Zhengyou calibration method to obtain internal parameters and external parameters of the binocular camera; S200: Based on the SURF algorithm and the region growing method, the same-name points of the target personnel are found in the image data. Specifically, the SURF algorithm is used to extract the feature points in the image data, and then the region growing is performed in the neighborhood of the feature points. The matching is further screened according to the grayscale similarity and geometric consistency of the region to obtain the coordinates of the matched same-name points. S300: Based on the principle of triangulation, according to the internal and external parameters of the binocular camera and the coordinates of the matched points of the same name, the position coordinates of the personnel target in the three-dimensional space are calculated. The specific calculation method is: Assume the optical center of the left camera is , the optical center of the right camera is , the coordinates of the points with the same name in the left image data are , the coordinates in the right image data are , the baseline length is The focal length of the camera is , then the target point Coordinates in three-dimensional space The specific calculation formulas are: , and in, and is the coordinate of the left camera principal point, For parallax.

[0012] Preferably, based on the position coordinates of the human target in the three-dimensional space, the adaptive motion model is established as follows: For speed changes less than And the change in movement direction is less than The uniform linear motion target is predicted and tracked based on the Kalman filter algorithm. The state vector of the Kalman filter is defined as ,in is the position coordinate of the target on the two-dimensional plane, For the purpose and The velocity component in the direction of and the observation equation Perform forecast updates; in, For the moment, , Respectively represent the target at time and time The state vector of is the state transfer matrix, is the process noise vector, Expressed as time The observation vector, is the observation matrix, is the observation noise vector; Based on the continuously predicted and updated position coordinate information, an adaptive motion model is established.

[0013] A visual sensing personnel positioning and tracking method comprises the following steps: The image data of personnel are collected by multiple binocular cameras distributed in the monitoring area, and the image data are transmitted to the data preprocessing module through a high-speed data transmission channel according to the preset frame rate and resolution; The received image data is processed in sequence by grayscale conversion, filtering and denoising, and histogram equalization; The target detection algorithm is used to detect the human targets in the processed image, and the LBP texture features and the color features after PCA dimensionality reduction of each target are extracted, and the information is passed to the positioning calculation and tracking module; The position coordinates of the personnel target in three-dimensional space are calculated through binocular vision principles and stereo vision algorithms, and an adaptive motion model is established based on the position coordinates.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The positioning calculation and tracking module of the present invention establishes an adaptive motion model according to the motion mode of the target, can accurately predict and track the position change of the target, effectively respond to various motion situations of the personnel, ensure the continuity and accuracy of tracking, and avoid target loss. When establishing the motion model, the present invention converts the color image into a grayscale image, retains only the brightness information, and removes the interference of the color channel, which not only reduces the amount of data and reduces the complexity of subsequent processing, but also makes the image more resistant to noise in some cases, and is not affected by color noise. It can effectively remove interference factors such as noise points and spots in the image, such as Gaussian noise, salt and pepper noise, etc., make the image clearer, reduce the interference of noise on the characteristics of the personnel target in the image, and improve the accuracy of subsequent target detection and feature extraction. In addition, by adjusting the grayscale distribution of the image, the grayscale histogram of the original image is changed from a relatively concentrated grayscale interval to a uniform distribution within the entire grayscale range, effectively expanding the dynamic range of the image and enhancing the global contrast of the image, especially for images with too bright or too dark backgrounds and foregrounds, the effect is significant, and the difference between the personnel target and the background can be better highlighted. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of a visual sensing personnel positioning and tracking system of the present invention; Figure 2 The present invention provides a flow chart of a visual sensing personnel positioning and tracking method. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0017] Refer to the attached Figure 1 , a visual sensing personnel positioning and tracking system, comprising: The visual acquisition module consists of multiple binocular cameras with high resolution, high frame rate and different viewing angles distributed in the monitoring area. The binocular cameras can automatically adjust the acquisition parameters based on the ambient light intensity and transmit the collected image data to the data preprocessing module in real time through a high-speed data transmission interface; The data preprocessing module is used to grayscale, filter and remove noise, and perform histogram equalization on the received image data to improve the image quality and highlight the target features of the person; The target detection and feature extraction module detects human targets from the processed image data based on the target detection algorithm, and extracts the LBP texture features and the color features after PCA dimensionality reduction of each human target; The positioning calculation and tracking module uses binocular vision principles and stereo vision algorithms to calculate the position coordinates of personnel targets in three-dimensional space, and establishes an adaptive motion model based on the position coordinates.

[0018] In this embodiment, the binocular camera in the visual acquisition module has the function of automatically adjusting the focus, aperture and white balance, and the data transmission interface is a Gigabit Ethernet interface or a fiber optic interface.

[0019] Specifically, the grayscale processing in the data preprocessing module is based on the weighted average method, specifically, the values ​​of the three RGB channels of each pixel in the image data collected by the binocular camera are multiplied by preset weight coefficients, the preset weight coefficient of the red channel is 0.299, the preset weight coefficient of the green channel is 0.587, and the preset weight coefficient of the blue channel is 0.114. The weighted values ​​of the three channels are added to obtain the grayscale value of the pixel to reduce the amount of data and highlight the contour features of the target person.

[0020] Specifically, for any pixel in the image , where the original The values ​​are recorded as , and , obtain the gray value of the pixel through weighted calculation , Specifically: ; According to the above formula, each pixel in the image is calculated in turn, and the obtained grayscale value is reassigned to the corresponding pixel position, so as to convert the entire color image into a grayscale image. This weighted average grayscale conversion method can effectively reduce the amount of data and the computational complexity of subsequent processing while retaining the main brightness and contrast information of the image. In addition, since some irrelevant details and interference factors in the color information are removed, the contour features of the human target are more prominent in the grayscale image, providing a clearer and more concise data basis for subsequent processing steps such as filtering, denoising and target detection.

[0021] In addition, the filtering and denoising steps in the data preprocessing module are as follows: S1: Perform median filtering on the grayscale image data. The median filter uses a window size of 3x3 pixels. Starting from the upper left corner of the image data, the image data is traversed pixel by pixel in units of windows. For each pixel point, all pixel grayscale values ​​in its neighborhood window are collected and sorted. The sorted middle value is used as the median filter grayscale value of the pixel point after median filtering to remove salt and pepper noise. S2: Gaussian filtering is performed on the image data after median filtering. The standard deviation of Gaussian filtering is set according to the noise situation of the image data and the required smoothness. The standard deviation is between 0.5-2, and the convolution kernel size is 6σ+1. The Gaussian kernel generated by the two-dimensional Gaussian distribution function is used to perform convolution operation on the image data. Specifically, the center of the Gaussian kernel is aligned with each pixel in the image data, and the weighted sum of the Gaussian kernel and the pixels in the neighborhood of the pixel is calculated. The weighted sum is used as the Gaussian filter grayscale value to remove Gaussian noise and retain the basic edge and contour information of the person target to avoid excessive blurring of the target edge.

[0022] It should be noted that, in step S1: First, determine the window size of the median filter to be 3X3 pixels. Starting from the upper left corner pixel of the image data, take the 3X3 window as the basic unit, and traverse the entire image data pixel by pixel from left to right and from top to bottom. For each pixel to be processed , extract the grayscale values ​​of all pixels in the surrounding 3X3 neighborhood to form a one-dimensional array , for the array To sort the elements in, you can use algorithms such as quick sort or bubble sort to get the sorted array ; Then the array The middle value in is taken as the pixel point The median filter grayscale value after median filtering replaces the original grayscale value. In this way, for the salt and pepper noise that may exist in the image data, since its grayscale value is greatly different from the grayscale value of the surrounding normal pixels, during the median filtering process, the grayscale value of these noise points will be replaced by a more reasonable grayscale value in the neighborhood, thereby effectively removing the salt and pepper noise, making the image data smoother and cleaner, and creating good conditions for the subsequent Gaussian filtering operation; It should be noted that, in step S2: First, the standard deviation of the Gaussian filter Set it between 0.5-2. For example, in a general indoor personnel positioning scenario, if the image data noise is relatively small, it can be set to; in some outdoor scenes with complex lighting and more noise, it can be set to Increase to about 1.5.

[0023] Based on the selected standard deviation , calculate the size of the Gaussian convolution kernel; Then, a Gaussian convolution kernel is generated. The Gaussian kernel is a matrix generated based on a two-dimensional Gaussian distribution function. The weight of the central pixel is the largest, and the weight gradually decreases as it moves away from the central pixel. For a size of ( is an odd number) of Gaussian convolution kernels , its position The weight value at , the specific calculation method is: in, is the Gaussian convolution kernel at coordinates The pixel weight value at the position, is the standard deviation of the Gaussian distribution, is the coordinate variable in the Gaussian convolution kernel.

[0024] Secondly, perform a Gaussian filter operation on the image data, align the center of the generated Gaussian convolution kernel with each pixel in the image data, and then for each pixel , calculate the weighted sum of the Gaussian kernel and the pixels in the neighborhood of the pixel as the Gaussian filtered gray value of the pixel after Gaussian filtering , The specific calculation is: ; in, Represents the pixel points in the image data The original gray value, is the Gaussian kernel at position The pixel weight value at .

[0025] By performing such a convolution operation on each pixel in the image data, Gaussian filtering can effectively remove Gaussian noise in the image data, making the image data smoother and more uniform, and reducing the interference of noise on subsequent target detection and feature extraction algorithms. At the same time, when appropriate parameters are selected, the basic edge and contour information of the human target can be better retained to avoid excessive blurring of the target's edge, thereby providing a more stable and reliable data basis for subsequent image data processing steps while ensuring the quality of the image data.

[0026] The combination of the two filtering methods can comprehensively improve the noise removal ability of image data. Median filtering is good at processing salt and pepper noise, while Gaussian filtering has a good inhibitory effect on Gaussian noise. They complement each other and can cope with complex noise environments. For example, in actual personnel positioning scenarios, image data may be interfered with by multiple types of noise at the same time. For example, in a shopping mall monitoring scenario, monitoring image data may generate salt and pepper noise due to signal transmission problems, and Gaussian noise due to factors such as light fluctuations in the environment. Using median filtering to remove salt and pepper noise first, and then using Gaussian filtering to remove Gaussian noise, can make image data clearer and cleaner, providing high-quality image data for subsequent operations such as target detection and feature extraction.

[0027] Furthermore, the histogram equalization processing in the data preprocessing module is specifically to perform histogram equalization processing on the image data after filtering and denoising, so as to make the grayscale value distribution of the image data more uniform and enhance the contrast of the image data; Specifically, the grayscale value range of the filtered and denoised image data is divided into 256 sub-intervals, and the number of pixels in each sub-interval is counted to obtain a grayscale value histogram. The cumulative distribution function of each gray level is calculated based on the grayscale value histogram. , for each gray level, The specific calculation method is: ; in, is the new gray value, is the cumulative distribution function value of the current gray level, is the minimum value in the cumulative distribution function, are the number of rows and columns of image data respectively.

[0028] It should be noted that It is the minimum value in the cumulative distribution function and is used for normalization when calculating the new grayscale value to ensure that the new grayscale value ranges from 0 to 255. It should also be noted that there are 256 grayscale values ​​in the grayscale value range from 0 to 255, that is, 0, ..., 255, a total of 256 grayscale values.

[0029] In this way, the area with concentrated grayscale value distribution in the original image data is stretched, so that the grayscale value distribution of the image data is more uniform, the contrast of the image data is enhanced, and the characteristics of the personnel target in the image data are more obvious, and the distinction from the background is greater, which is convenient for subsequent target detection and feature extraction operations, and effectively improves the recognition and positioning capabilities of the entire system for personnel targets. In summary, combined with the above, in the visual sensing personnel positioning and tracking system, the visual acquisition module will obtain a series of image data containing personnel. For each such image data, it must first be grayed, then filtered and denoised, and finally histogram equalization processing flow is performed to ensure that each image data used for subsequent target detection and feature extraction operations has high quality, that is, clear personnel target contours, low noise interference and good contrast.

[0030] In this embodiment, the positioning calculation and tracking module uses binocular vision principle and stereo vision algorithm to calculate the position coordinates of the personnel target in three-dimensional space. The camera in the visual acquisition module is a binocular camera, which specifically includes the following steps: S100: The binocular camera uses Zhang Zhengyou calibration method to obtain the internal and external parameters of the binocular camera; S200: Based on the SURF algorithm and the region growing method, the same-name points of the target personnel are found in the image data. Specifically, the SURF algorithm is used to extract the feature points in the image data, and then the region growing is performed in the neighborhood of the feature points. The matching is further screened according to the grayscale similarity and geometric consistency of the region to obtain the coordinates of the matched same-name points. It should be noted that in the context of binocular vision systems being used for personnel positioning and tracking, homonymous points refer to the pixel points that represent the same actual physical point in space (here, a point on the target person) in image data from different perspectives (for example, image data captured by the left and right cameras in binocular vision). In simple terms, it is the correspondence between the pixel points at the same position on the person in different image data when the same person is captured from different angles.

[0031] S300: Based on the principle of triangulation, according to the internal and external parameters of the binocular camera and the coordinates of the matched points of the same name, the position coordinates of the personnel target in the three-dimensional space are calculated. The specific calculation method is: Assume the optical center of the left camera is , the optical center of the right camera is , the coordinates of the points with the same name in the left image data are , the coordinates in the right image data are , the baseline length is , the focal length of the left camera is , then the target point Coordinates in three-dimensional space The specific calculation formulas are: , and in, and is the coordinate of the left camera principal point, For parallax.

[0032] Through this precise triangulation calculation, the position information of personnel targets in three-dimensional space can be accurately obtained, providing a high-precision position data basis for subsequent tracking and positioning, and effectively improving the system's spatial positioning accuracy and reliability for personnel targets.

[0033] It should also be noted that based on the position coordinates of the personnel target in the three-dimensional space, the adaptive motion model is established as follows: For speed changes less than And the change in movement direction is less than The uniform linear motion target is predicted and tracked based on the Kalman filter algorithm. The state vector of the Kalman filter is defined as ,in is the position coordinate of the target on the two-dimensional plane, For the purpose and The velocity component in the direction of and the observation equation Perform forecast updates; in, For the moment, , Respectively represent the target at time and time The state vector of is the state transfer matrix, is the process noise vector, Expressed as time The observation vector, is the observation matrix, is the observation noise vector; Based on the continuously predicted and updated position coordinate information, an adaptive motion model is established.

[0034] Through continuous prediction and updating steps, and using the historical position observations and current observations of the target, the Kalman filter can effectively estimate the current position and speed of the target, achieve stable tracking of the target in uniform linear motion, and reduce the computational complexity while ensuring tracking accuracy, thereby improving the real-time performance and efficiency of the system.

[0035] Refer to the attached Figure 2 , a visual sensing personnel positioning and tracking method, comprising the following steps: The image data of personnel are collected by multiple binocular cameras distributed in the monitoring area, and the image data are transmitted to the data preprocessing module through a high-speed data transmission channel according to the preset frame rate and resolution; The received image data is processed in sequence by grayscale conversion, filtering and denoising, and histogram equalization; The target detection algorithm is used to detect the human targets in the processed image, and the LBP texture features and the color features after PCA dimensionality reduction of each target are extracted, and the information is passed to the positioning calculation and tracking module; The position coordinates of the personnel target in three-dimensional space are calculated through binocular vision principles and stereo vision algorithms, and an adaptive motion model is established based on the position coordinates.

[0036] In the description of the present invention, it should be understood that the above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0037] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0038] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or multiple.

[0039] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0040] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0041] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0042] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0043] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0044] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0045] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage media include: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks or optical disks.

[0046] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A visual sensing personnel positioning and tracking system, characterized in that: include: The visual acquisition module consists of multiple binocular cameras with high resolution, high frame rate and different viewing angles distributed in the monitoring area. The binocular cameras can automatically adjust the acquisition parameters based on the ambient light intensity and transmit the collected image data to the data preprocessing module in real time through a high-speed data transmission interface; The data preprocessing module is used to perform grayscale conversion, filtering and denoising, and histogram equalization processing on the received image data; The target detection and feature extraction module detects human targets from the processed image data based on the target detection algorithm, and extracts the LBP texture features and the color features after PCA dimensionality reduction of each human target; The positioning calculation and tracking module uses binocular vision principles and stereo vision algorithms to calculate the position coordinates of personnel targets in three-dimensional space, and establishes an adaptive motion model based on the position coordinates.

2. The visual sensing personnel positioning and tracking system according to claim 1, characterized in that: The binocular camera in the visual acquisition module has the function of automatically adjusting focus, aperture and white balance, and the data transmission interface is a Gigabit Ethernet interface or an optical fiber interface.

3. The visual sensing personnel positioning and tracking system according to claim 1, characterized in that: The grayscale processing in the data preprocessing module is based on the weighted average method. Specifically, the values ​​of the three RGB channels of each pixel in the image data collected by the binocular camera are multiplied by preset weight coefficients, the preset weight coefficient of the red channel is 0.299, the preset weight coefficient of the green channel is 0.587, and the preset weight coefficient of the blue channel is 0.

114. The weighted values ​​of the three channels are added to obtain the grayscale value of the pixel.

4. The visual sensing personnel positioning and tracking system according to claim 3, characterized in that: The filtering and denoising steps in the data preprocessing module are as follows: S1: Perform median filtering on the grayscaled image data. The median filter uses a window size of 3x3 pixels. Starting from the upper left corner of the image data, the image data is traversed pixel by pixel in units of the window. For each pixel point, all pixel grayscale values ​​in its neighborhood window are collected and sorted, and the sorted middle value is used as the median filtered grayscale value of the pixel point after median filtering. S2: Perform Gaussian filtering on the image data after median filtering. The standard deviation of Gaussian filtering is set according to the noise situation of the image data and the required smoothness. The standard deviation is between 0.5-2, and the convolution kernel size is 6σ+1. The Gaussian kernel generated by the two-dimensional Gaussian distribution function is used to perform convolution operation on the image data. Specifically, the center of the Gaussian kernel is aligned with each pixel in the image data, and the weighted sum of the Gaussian kernel and the pixels in the neighborhood of the pixel is calculated. The weighted sum is used as the Gaussian filter gray value.

5. The visual sensing personnel positioning and tracking system according to claim 4, characterized in that: The histogram equalization processing in the data preprocessing module is specifically to perform histogram equalization processing on the image data after filtering and denoising; Specifically, the grayscale value range of the filtered and denoised image data is divided into 256 sub-intervals, and the number of pixels in each sub-interval is counted to obtain a grayscale value histogram. The cumulative distribution function of each gray level is calculated based on the grayscale value histogram. , for each gray level, The specific calculation method is: ； in, is the new gray value, is the cumulative distribution function value of the current gray level, is the minimum value in the cumulative distribution function, are the number of rows and columns of image data respectively.

6. The visual sensing personnel positioning and tracking system according to claim 1, characterized in that: The positioning calculation and tracking module adopts binocular vision principle and stereo vision algorithm to calculate the position coordinates of the personnel target in three-dimensional space. The camera in the visual acquisition module is a binocular camera, which specifically includes the following steps: S100: The binocular camera uses the Zhang Zhengyou calibration method to obtain internal parameters and external parameters of the binocular camera; S200: Based on the SURF algorithm and the region growing method, the same-name points of the target personnel are found in the image data. Specifically, the SURF algorithm is used to extract the feature points in the image data, and then the region growing is performed in the neighborhood of the feature points. The matching is further screened according to the grayscale similarity and geometric consistency of the region to obtain the coordinates of the matched same-name points. S300: Based on the principle of triangulation, according to the internal and external parameters of the binocular camera and the coordinates of the matched points of the same name, the position coordinates of the personnel target in the three-dimensional space are calculated. The specific calculation method is: Assume the optical center of the left camera is , the optical center of the right camera is , the coordinates of the points with the same name in the left image data are , the coordinates in the right image data are , the baseline length is , the focal length of the left camera is , then the target point Coordinates in three-dimensional space The specific calculation formulas are: , and in, and is the coordinate of the left camera principal point, For parallax.

7. The visual sensing personnel positioning and tracking system according to claim 6, characterized in that: Based on the position coordinates of the personnel target in three-dimensional space, the adaptive motion model is established as follows: For speed changes less than And the change in movement direction is less than The uniform linear motion target is predicted and tracked based on the Kalman filter algorithm. The state vector of the Kalman filter is defined as ,in is the position coordinate of the target on the two-dimensional plane, For the purpose and The velocity component in the direction of and the observation equation Perform forecast updates; in, For the moment, , Respectively represent the target at time and time The state vector of is the state transfer matrix, is the process noise vector, Expressed as time The observation vector, is the observation matrix, is the observation noise vector; Based on the continuously predicted and updated position coordinate information, an adaptive motion model is established.

8. A visual sensing personnel positioning and tracking method, applied to the visual sensing personnel positioning and tracking system according to any one of claims 1 to 7, characterized in that: The steps include: The image data of personnel are collected by multiple binocular cameras distributed in the monitoring area, and the image data are transmitted to the data preprocessing module through a high-speed data transmission channel according to the preset frame rate and resolution; The received image data is processed in sequence by grayscale conversion, filtering and denoising, and histogram equalization; The target detection algorithm is used to detect the human targets in the processed image, and the LBP texture features and the color features after PCA dimensionality reduction of each target are extracted, and the information is passed to the positioning calculation and tracking module; The position coordinates of the personnel target in three-dimensional space are calculated through binocular vision principles and stereo vision algorithms, and an adaptive motion model is established based on the position coordinates.

Images