A semi-supervised artificial intelligence human body detection embedded system based on array timing

Abstract

The present invention discloses a semi-supervised artificial intelligence human body detection embedded system based on array timing, which adopts a modular programming method. The system includes seven modules, namely, a human body numbering module, a sequence production module, a trend analysis module, a result comparison module, a verification module and an error analysis module, and a human body detection module, which are connected in sequence. The present invention can predict the number of people in a specific time period according to the historical flow of people in a specified area, which is conducive to the staff to formulate strategies in advance according to the prediction results. In the process of human body detection, the present invention adopts pixel gradient vector features for comparison and identification. Through verification, it is found that the system effectively improves the recognition accuracy.

Description

Technical Field

[0001] The invention relates to a semi-supervised artificial intelligence human body detection embedded system based on array timing, belonging to the technical field of monitoring systems. Background Art

[0002] Embedded supervision systems are widely used, mainly through monitoring equipment for supervision and management. For some places, when it is necessary to count the flow of people, the existing supervision and detection system identifies and counts the human body in the image to ensure that the statistical purpose is achieved, thereby making a strategic reservation for the flow of people in the place. However, there are disadvantages in the use of the existing system. It does not have the function of predicting based on the counted flow of people. The function is single, which is not conducive to assisting staff to pre-make response strategies for crowd control. Secondly, the human body contour is often detected based on grayscale and polarization. When the human body is together, the contour detection will overlap, which will cause a large deviation in the counting of the human body. Therefore, the effect of human body recognition detection is poor and the accuracy is not high. In addition, the existing system cannot provide feedback on the errors after the inspection, and the staff cannot update the system and adjust the parameters in time, resulting in poor use effect. Summary of the invention

[0003] In view of the problems existing in the above-mentioned prior art, the present invention provides a semi-supervised artificial intelligence human body detection embedded system based on array timing, so as to solve the above-mentioned technical problems.

[0004] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a semi-supervised artificial intelligence human body detection embedded system based on array timing, including a human body numbering module, a sequence production module, a trend analysis module, a result comparison module, a verification module and an error analysis module connected in sequence, and also including a human body detection module,

[0005] Among them: the human body numbering module numbers the data of human bodies of different body shapes; the sequence generation module is used to generate the matrix time series of human body appearance; the trend analysis module is used to predict the matrix time series and predict the time period when the human body will appear next; the trend analysis module uses the number statistical prediction and interval statistical prediction method to make two predictions; the result comparison module compares the two predictions made by the trend analysis module; the verification module compares and verifies the time period of human body appearance detected by the human body detection module with the prediction; the error analysis module is used to analyze the reasons for the errors in the comparison.

[0006] Furthermore, the human body numbering module is connected to a human body classification module, and the human body classification module classifies the human body into an adult or a child according to the human body's height.

[0007] Furthermore, the human body detection module includes an image acquisition device, a human body detection module and a detection result output module, wherein: the image acquisition device is used to monitor human body images in the shooting area; the human body detection module is used to detect human bodies in the image; and the result output module is used to output human body detection results.

[0008] Furthermore, the human body detection module includes a recognition module and a comparison module, the recognition module is used to recognize the human body contour in the image, and the comparison module is used to compare the recognized human body contour with the stored human body contour data.

[0009] Furthermore, the recognition module includes a grayscale module, a normalization module, a pixel gradient module, a grid module, a gradient histogram generation module, a grouping module, a group series module and a feature vector group output module which are connected in sequence, wherein: the grayscale module grayscales the image; the normalization module uses the Gamma correction method to standardize the color space of the input image; the pixel gradient module calculates the gradient of each pixel, and the gradient includes direction and size; the grid module divides a number of adjacent pixels into square grids; the gradient histogram generation module generates a gradient histogram of each grid; the grouping module groups adjacent grids; the group series module combines the groups to obtain the image feature vector and outputs it by the feature vector group output module.

[0010] Furthermore, the trend analysis module includes a frequency statistics module and an interval statistics module. The frequency statistics module is used to count the time periods and the number of human bodies that appear, and the interval statistics module is used to count the total number of human bodies within the total time period and reversely infer the time periods when human bodies appear by counting the intervals of time periods when human bodies do not appear. The frequency statistics module and the interval statistics module are respectively connected to a mean module, and the mean module is connected to a prediction module. The mean module is used to calculate the mean of the time period, and then predict the time period when human bodies appear through the prediction module.

[0011] Furthermore, it also includes a human body database, which is used to store and update human body data. The human body database includes library update and library input. The library update is used to update the existing human body data in the human body database, and the library input is used to add human body data to the human body database.

[0012] Furthermore, the sequence generation module includes a data loading module, an endpoint date extraction module, a continuous time series generation module, a data bundling module and a matrix drawing module. The data loading module is used to load human body data, the endpoint date extraction module records the starting node of the time period, and then generates a continuous sequence at two nodes through the continuous time series generation module. The data bundling module arranges the human body data into the corresponding sequence, and the matrix drawing module generates and sends out a matrix time series.

[0013] The beneficial effects of the present invention are:

[0014] 1. Through the set sequence generation module, the data loading module is used to load human body data, the endpoint date extraction module records the starting node of the time period, and then the continuous time series generation module generates a continuous sequence at two nodes. The data bundling module arranges the human body data into the corresponding sequence, and the matrix drawing module generates and issues a matrix time series, and then cooperates with the mean module to calculate the mean of the time period. After that, the prediction module predicts the time period when the human body appears. Therefore, according to the obtained human body counting data, after generating the matrix time series, the number of human bodies that may appear in the target time period is deduced according to the matrix time series analysis, which is conducive to providing reference for staff and formulating strategies in advance.

[0015] 2. Through the set trend analysis module, the frequency statistics module is used to count the time periods and the number of human bodies when the human body appears, the interval statistics module is used to count the total number of human bodies in the total time period and reversely infer the time period when the human body appears by counting the intervals of the time periods when the human body does not appear. The frequency statistics module and the interval statistics module are respectively connected to the mean module, and the mean module is connected to the prediction module. The mean module is used to calculate the mean of the time period, and then the prediction module is used to predict the time period when the human body appears. Two methods are used to predict the number of human bodies, which improves the prediction accuracy.

[0016] 3. After the number of human bodies in the area is predicted, the verification module is used to compare and verify the time period of human bodies detected by the human detection module with the prediction, and then the error analysis module is used to analyze the reasons for the errors in the comparison. The errors in the entire system can be well analyzed and fed back, and the system can be adjusted by the staff to ensure the long-term use of the system and guarantee accuracy.

[0017] 4. The human body images in the shooting area are monitored by the set image acquisition device. The human body detection module is used to detect the human body in the image, the result output module is used to output the human body detection result, the recognition module is used to identify the human body feature vector in the image, and the comparison module is used to compare the recognized human body feature vector with the stored human body feature vector data. Specifically, in the recognition module, the grayscale module grayscales the image, the normalization module uses the Gamma correction method to standardize the color space of the input image, the pixel gradient module calculates the gradient of each pixel, and the gradient includes direction and size. The grid module divides several adjacent pixels into square grids, the gradient histogram generation module generates the gradient histogram of each grid, the grouping module groups the adjacent grids, and the group series module combines the groups to obtain the image feature vector and outputs it by the feature vector group output module. The traditional human body contour detection and recognition method is changed to the human body feature vector recognition method, which effectively improves the accuracy of human body detection and recognition and improves the counting accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the semi-supervised artificial intelligence human body detection embedded system based on array timing of the present invention;

[0019] Figure 2 It is a schematic diagram of a sequence generation module of an embedded system for semi-supervised artificial intelligence human body detection based on array timing of the present invention;

[0020] Figure 3 It is a schematic diagram of a trend analysis module of the semi-supervised artificial intelligence human body detection embedded system based on array timing of the present invention;

[0021] Figure 4 It is a schematic diagram of the recognition module of the semi-supervised artificial intelligence human body detection embedded system based on array timing of the present invention;

[0022] Figure 5 This is a schematic diagram of the human body database of the semi-supervised artificial intelligence human body detection embedded system based on array timing of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention more clear, the present invention is further described in detail below through the accompanying drawings and embodiments. However, it should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the scope of the present invention.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0025] like Figure 1-Figure 5 As shown, a semi-supervised artificial intelligence human body detection embedded system based on array timing includes a human body numbering module, a sequence production module, a trend analysis module, a result comparison module, a verification module and an error analysis module connected in sequence, and also includes a human body detection module.

[0026] Among them: the human body numbering module numbers the data of human bodies of different body shapes; the sequence generation module is used to generate the matrix time series of human body appearance; the trend analysis module is used to predict the matrix time series and predict the time period when the human body will appear next; the trend analysis module uses the number statistical prediction and interval statistical prediction method to make two predictions; the result comparison module compares the two predictions made by the trend analysis module; the verification module compares and verifies the time period of human body appearance detected by the human body detection module with the prediction; the error analysis module is used to analyze the reasons for the errors in the comparison.

[0027] In this embodiment, the human body numbering module is preferably connected to a human body classification module, and the human body classification module classifies the human body into an adult or a child according to the human body's height. Artificial numbering can be performed, such as 1 represents an adult and 2 represents a minor. The height demarcation limit can be set as needed, such as setting the height of 1.2m as the limit, if it is higher, it is recorded as 1, and if it is lower, it is recorded as 2, which can assist the sequence generation module in generating the time sequence.

[0028] In this embodiment, the human body detection module preferably includes an image acquisition device, a human body detection module and a detection result output module, wherein: the image acquisition device is used to monitor the human body image in the shooting area; the human body detection module is used to detect the human body in the image; that is, the part where the human body exists in the image transmitted by the image acquisition device is captured and compared. The capture and comparison of the human body detection module are disclosed in detail below: the result output module is used to output the human body detection result.

[0029] Preferably, in this embodiment, the human body detection module includes a recognition module and a comparison module, the recognition module is used to recognize the human body contour in the image, and the comparison module is used to compare the recognized human body contour with the stored human body contour data.

[0030] In this embodiment, the recognition module preferably includes a grayscale module, a normalization module, a pixel gradient module, a grid module, a gradient histogram generation module, a grouping module, a group series module and a feature vector group output module which are connected in sequence, wherein: the grayscale module grayscales the image; the normalization module uses the Gamma correction method to standardize the color space of the input image; the pixel gradient module calculates the gradient of each pixel, and the gradient includes direction and size; the grid module divides a number of adjacent pixels into square grids; the gradient histogram generation module generates a gradient histogram of each grid; the grouping module groups adjacent grids; the group series module combines the groups to obtain the image feature vector and outputs it by the feature vector group output module.

[0031] The pixel gradient part can be performed using the following OpenCV-based algorithm:

[0032] #include <iostream>

[0033] #include<cv.h>

[0034] #include<opencv2 / core / core.hpp>

[0035] #include<opencv2 / highgui / highgui.hpp>

[0036] #include<opencv2 / imgproc / imgproc.hpp>

[0037] using namespace cv;

[0038] using namespace std;

[0039] int main()

[0040] Mat picture = imread("test.jpg", 0); / / Grayscale

[0041] Mat img;

[0042] picture.convertTo(img, CV_32F); / / Convert to floating point

[0043] sqrt(img, img); / / gamma correction

[0044] normalize(img, img, 0, 255, NORM_MINMAX, CV_32F); / / Normalize [0,255] floating point number

[0045] Mat gradient = Mat::zeros(img.rows, img.cols, CV_32F); / / Gradient

[0046] Mat theta = Mat::zeros(img.rows, img.cols, CV_32F); / / Angle

[0047] for (int i = 1; i < img.rows - 1; i++)

[0048] {

[0049] for (int j = 1; j < img.cols - 1; j++)

[0050] {

[0051] float Gx, Gy;

[0052] Gx = img.at <float>(i, j + 1) - img.at <float>(i, j - 1);

[0053] Gy = img.at <float>(i + 1, j) - img.at <float>(i - 1, j);

[0054] gradient.at <float>(i, j) = sqrt(Gx * Gx + Gy * Gy); / / Gradient modulus

[0055] theta.at <float>(i, j) = float(atan2(Gy, Gx) * 180 / CV_PI); / / gradient direction [-180°, 180°]

[0056] }

[0057] }

[0058] normalize(gradient, gradient, 0, 255, NORM_MINMAX, CV_8UC1); / / Normalization [0,255] unsigned integer

[0059] normalize(img, img, 0, 255, NORM_MINMAX, CV_8UC1);

[0060] imshow("gradient map", gradient);

[0061] waitKey();

[0062] return 0;

[0063] }

[0064] In the above algorithm, the horizontal direction of the gradient is: [-1 0 1], and the vertical direction is: [-1 0 1] T , the final output result is the gradient map (that is, the gradient finally obtained in the algorithm).

[0065] In this embodiment, the trend analysis module preferably includes a frequency statistics module and an interval statistics module. The frequency statistics module is used to count the time period and number of human bodies that appear in the human body. The interval statistics module is used to count the total number of human bodies in the total time period and reversely infer the time period in which the human body appears by counting the interval of the time period in which the human body does not appear. The frequency statistics module and the interval statistics module are respectively connected to the mean module, and the mean module is connected to the prediction module. The mean module is used to calculate the mean of the time period, and then the prediction of the time period in which the human body appears is performed through the prediction module. For example, for the frequency statistics module: if the number of people appearing in the time period from 9:00 to 10:00 is 10, it is inferred that the number of people appearing in the next hour is 10; for the interval statistics module: if the total number of people appearing from 9:00 to 11:00 is 10, among which no person appears from 10:00 to 11:00, the mean module infers that no person will appear in the next half hour after averaging, and the number of people who will appear in the next half hour time period is 5. In this way, it is inferred that the number of people appearing in the next hour is 5. Obviously, the two inference methods have different prediction results.

[0066] Preferably, the present embodiment further includes a human body database, which is used to store and update human body data. The human body database includes library update and library input. The library update is used to update the existing human body data in the human body database, and the library input is used to add human body data to the human body database.

[0067] In this embodiment, the sequence generation module preferably includes a data loading module, an endpoint date extraction module, a continuous time series generation module, a data bundling module and a matrix drawing module. The data loading module is used to load human body data. The endpoint date extraction module records the starting node of the time period, and then generates a continuous sequence at two nodes through the continuous time series generation module. The data bundling module arranges the human body data into the corresponding sequence. The matrix drawing module generates and sends out a matrix time series. The data bundling module arranges the human body data into the corresponding sequence. The corresponding data (recorded as 1 or 2) for identifying adults or children according to height can be bound at the same time. The matrix drawing module generates and sends out a matrix time series.

[0068] The present invention sets a sequence generation module, a data loading module is used to load human body data, an endpoint date extraction module records the starting node of a time period, and then generates a continuous sequence at two nodes through a continuous time series generation module. The data bundling module arranges the human body data into a corresponding sequence, and a matrix drawing module generates and issues a matrix time series. The mean module is then used to calculate the mean of the time period, and then the prediction module is used to predict the time period of human body appearance. Therefore, according to the obtained human body counting data, after the matrix time series is generated, the number of human bodies that may appear in the target time period is derived according to the matrix time series analysis, which is beneficial to provide staff with Reference, formulate strategies in advance; through the set trend analysis module, the frequency statistics module is used to count the time period and number of human bodies, the interval statistics module is used to count the total number of human bodies in the total time period and reversely infer the time period when the human body appears by counting the intervals of the time period when the human body does not appear, the frequency statistics module and the interval statistics module are respectively connected to the mean module, the mean module is connected to the prediction module, the mean module is used to calculate the mean of the time period, and then the prediction module is used to predict the time period when the human body appears. Two methods are used to predict the number of human bodies, which improves the prediction accuracy; after the number of human bodies in the area is predicted, the time period of human body appearance detected by the human body detection module is compared and verified with the prediction through the verification module, and then the error analysis module is used to analyze the reasons for the errors in the comparison. The errors existing in the entire system can be well analyzed and fed back, and then the system can be adjusted by the staff to ensure the long-term use of the system and ensure accuracy; the human body images in the shooting area are monitored by the set image acquisition device, the human body detection module is used to detect the human body in the image, the result output module is used to output the human body detection result, the recognition module is used to identify the human body feature vector in the image, and the comparison module is used to compare the recognized human body feature vector with the stored human body feature vector data, specifically In the recognition module, the grayscale module grayscales the image, the normalization module uses the Gamma correction method to standardize the color space of the input image, the pixel gradient module calculates the gradient of each pixel, and the gradient includes direction and size. The grid module divides several adjacent pixels into square grids, the gradient histogram generation module generates the gradient histogram of each grid, the grouping module groups the adjacent grids, and the group series module combines the groups to obtain the image feature vector and outputs it by the feature vector group output module. The traditional human contour detection and recognition method is changed to the human feature vector recognition method, which effectively improves the accuracy of human detection and recognition and improves the counting accuracy.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent substitution or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.< / float> < / float> < / float> < / float> < / float> < / float> < / iostream>

Claims

1. A semi-supervised artificial intelligence human body detection embedded system based on array timing, characterized in that: It includes a human body numbering module, a sequence production module, a trend analysis module, a result comparison module, a verification module and an error analysis module connected in sequence, and also includes a human body detection module. Among them: the human body numbering module is used to number the data of human bodies of different body types; the sequence generation module is used to generate the matrix time sequence of human body appearance; the trend analysis module is used to predict the matrix time sequence and predict the time period of the next appearance of the human body; the trend analysis module uses the number of statistical predictions and interval statistical predictions to make two predictions; the result comparison module compares the two predictions made by the trend analysis module; the verification module compares and verifies the time period of human body appearance detected by the human body detection module with the prediction; the error analysis module is used to analyze the reasons for the errors in the comparison; The human body detection module includes an image acquisition device, a human body detection module and a detection result output module, wherein: the image acquisition device is used to monitor the human body image in the shooting area; the human body detection module is used to detect the human body in the image; and the result output module is used to output the human body detection result; The trend analysis module includes a frequency statistics module and an interval statistics module, the frequency statistics module is used to count the time period and number of human bodies that appear, the interval statistics module is used to count the total number of human bodies in the total time period and reversely infer the time period when the human body appears by counting the intervals of the time period when the human body does not appear, the frequency statistics module and the interval statistics module are respectively connected to a mean module, the mean module is connected to a prediction module, the mean module is used to calculate the mean of the time period, and then predict the time period when the human body appears through the prediction module; The sequence generation module includes a data loading module, an endpoint date extraction module, a continuous time series generation module, a data bundling module and a matrix drawing module. The data loading module is used to load human body data. The endpoint date extraction module records the starting node of the time period, and then generates a continuous sequence at two nodes through the continuous time series generation module. The data bundling module arranges the human body data into the corresponding sequence, and the matrix drawing module generates and sends out a matrix time series.

2. According to claim 1, a semi-supervised artificial intelligence human body detection embedded system based on array timing is characterized in that: The human body numbering module is connected to a human body classification module, and the human body classification module classifies the human body into an adult or a child according to the height of the human body.

3. According to the semi-supervised artificial intelligence human body detection embedded system based on array timing according to claim 1, it is characterized in that: The human body detection module includes a recognition module and a comparison module. The recognition module is used to recognize the human body contour in the image, and the comparison module is used to compare the recognized human body contour with the stored human body contour data.

4. According to claim 3, a semi-supervised artificial intelligence human body detection embedded system based on array timing is characterized in that: The recognition module comprises a grayscale module, a normalization module, a pixel gradient module, a grid module, a gradient histogram generation module, a grouping module, a group series module and a feature vector group output module which are connected in sequence, wherein: the grayscale module grayscales the image; the normalization module uses the Gamma correction method to standardize the color space of the input image; the pixel gradient module calculates the gradient of each pixel, and the gradient includes direction and size; the grid module divides a number of adjacent pixels into square grids; the gradient histogram generation module generates a gradient histogram of each grid; the grouping module groups adjacent grids; the group series module combines the groups to obtain an image feature vector and outputs it by the feature vector group output module.

5. According to claim 1, a semi-supervised artificial intelligence human body detection embedded system based on array timing is characterized in that: It also includes a human body database, which is used to store and update human body data. The human body database includes library update and library input. The library update is used to update the existing human body data in the human body database, and the library input is used to add human body data to the human body database.

Images