CN111415374A - KVM system and method for monitoring and managing scenic spot pedestrian flow - Google Patents

Abstract

The invention discloses a KVM system and method for monitoring and management of people flow in scenic spots, which is applied to an electronic device. The method performs moving object segmentation, moving object tracking and moving object counting; the system includes a processor FPGA chip and a computer program stored on the FPGA chip and executable on the processor, the computer program executing the method. The invention uses FPGA chips to perform parallel operations on video images, without processing speed and low delay, ensuring that changes in the flow of people in scenic spots can be detected in real time. In addition, the invention combines computer vision technology with KVM technology, effectively prolonging the period of time. The data transmission distance provides convenience for the monitoring and management of the flow of people in the scenic spot.

Description

A KVM system and method for monitoring and management of people flow in scenic spots

technical field

The present invention relates to the field of computer technology, in particular to a KVM method for monitoring and managing the flow of people in scenic spots.

Background technique

With the continuous improvement of social living standards and the continuous improvement of computer technology, for large public places such as shopping malls, stations, and tourist attractions, people counting has become an important task for management and decision-makers. A vital part of income. In the past, people used manual counting or manual electronic equipment to trigger counting. Obviously, this method can no longer adapt to the era of information explosion. There are many automatic counting methods, such as thermal imaging counting and infrared counting, but infrared counting. Counting is easily disturbed by external factors. For example, many people passing by will cause leakage and the instrument band will also affect it.

In order to improve the accuracy of people flow statistics, computer vision technology is introduced into it. It uses high-definition video acquisition equipment and intelligent digital image algorithms to accurately shoot, locate, track and count moving targets. The combination of visual technology and KVM technology effectively extends the data transmission distance and provides convenience for the monitoring and management of people flow in scenic spots.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to solve the deficiencies of the prior art. The present invention provides a KVM device and method for monitoring and managing the flow of people in scenic spots. A camera and a microphone are used to collect video and audio signals respectively, and an FPGA module is used. Perform video signal processing, and use the HiSilicon module to compress, encode and decode all data. Finally, the KVM device is used to realize the function of remote host computer to control different local video and audio, which solves the current lack of accurate statistics on the flow of people in the scenic area and for the scenic spot. issues of effective management.

In order to solve the aforementioned purpose, the technical solution of the present invention is a KVM method for monitoring and managing the flow of people in scenic spots, applied to an electronic device, and the method includes

S1: The three-frame difference method is used to detect the running target of the collected video image;

S2: Segment the running target of the video image after the running target detection has been completed;

S3: Take the tourist head area as the target feature, and set the enclosing rectangle of the second head area as the tracking window;

S4: Use Kalman filtering to predict the area of the next frame of the tracking window, and finally find the best matching object in the prediction area combined with the minimum Euclidean distance to achieve target tracking;

S5: When the next frame of image comes, if the tracked moving target prediction area is not within the shooting area, the count value will change, so as to count the running target.

The running target detection adopts the three-frame difference method, which includes the following steps:

S101: Use f _K-1 (x, y), f _K (x, y) and f _K+1 (x, y) to represent the k-1 th frame, the k th frame and the k+1 th frame image in (x , y) pixel grayscale of point;

S102: Differentiate f _K (x, y) and f _K-1 (x, y), make a difference between f _K+1 (x, y) and f _K (x, y), and record the difference image as D _K ( x,y), D _K+1 (x,y), D _K (x,y) are obtained by the following formulas:

S103: Perform AND operation on the difference images D _K (x, y) and D _K+1 (x, y) to obtain an image

S104: Use the maximum inter-class variance method to perform threshold segmentation to obtain a binary image

S105: Use the dilation operation to perform morphological processing, so that a clearer foreground moving target image Rn can be obtained.

Specifically, the operation target segmentation in the step S2 is characterized in that if the extracted moving target regions are adhered to each other, the segmentation process is performed, and if it is determined to be a single moving target, no segmentation is required, and the target segmentation step includes:

S201: Circumscribed rectangle of the connected domain: first, draw a circumscribed rectangle for the outer contour of the human body, and scan the rectangle to obtain the four-corner coordinates of the rectangle as X1n, X2n, Y1n, Y2n, where n is the sequence number of the sticking motion target rectangle;

S202: By the following formula:

D ₁ =|X _2N -X _1N |, D ₂ =|Y _2N -Y _1N |

Obtain the length and width of the circumscribed rectangle outline of the moving target, set the length threshold of the circumscribed rectangle as A, the width as B, and the aspect ratio as . If the ratio of length, width and length to width of the circumscribed rectangle is within the set threshold range, it is judged as a single moving target. If it is not within the range, it means that there is more than one moving target, and the circumscribed rectangle needs to be divided.

S203: adopt the vertical division method, calculate the area of each circumscribed rectangle and mark it as m ₀₀ and the two first-order moments are marked as m ₁₀ and m ₀₁ respectively;

得出质心坐标，x和y分别代 表质心的横坐标和纵坐标，质心个数则代表了游客目标人数，最后将质心 之间横坐标的均值作为标点，画垂直分割界限。S204: Using the centroid formula to calculate as

The centroid coordinates are obtained, x and y represent the abscissa and ordinate of the centroid, respectively, and the number of centroids represents the target number of tourists. Finally, the average of the abscissas between the centroids is used as a punctuation point to draw the vertical division boundary.

Further, the target tracking in the step S4 is characterized by comprising:

S301: Select target feature parameters;

S302: System model establishment and initialization;

S303: target prediction and matching: based on the center point of the current moving target, use Kalman filtering to predict the position of the center point of the next frame of moving target, when the next frame of image comes, set a pre-search area with a radius of R based on the predicted center point, Search with the minimum Euclidean distance within this range to obtain the best matching object;

S304: target update: take the best matching object as the initial position, update the target information, and repeat the above operation until the end.

Preferably, the selection of target feature parameters in the step S301 is characterized by comprising:

S401: First, use the Canny edge detection algorithm to extract the edge features of each rectangular frame;

S402: secondly utilize Hough transform to detect the circle in the upper half of the rectangular frame and use it as the head of the moving target;

S403: Make the detected head of the moving target a minimum circumscribed rectangle and scan the coordinates on the four corners of the circumscribed rectangle as X1n, X2n, Y1n, Y2n, where n is the sequence number of the moving target head, and calculate the target head window by the following formula length and width:

D ₁ =|X _2N -X _1N |, D ₂ =|Y _2N -Y _1N |

and by the formula:

Get the window center point (X,Y).

S404: Assuming that the pixel of a certain point in the circumscribed rectangle is f(x, y), and the total number of pixels in the head area is N, the average pixel value of the head area is

Specifically, the system model establishment and initialization in described step S302 is characterized in that comprising:

S501: The time difference between adjacent frames is very small, resulting in a small change in the speed v of the moving object, then it is determined that the moving object is moving at a uniform speed, and the center point of the window (X, Y), the pixel mean w and the speed v are selected as state variables;

S502: Calculate the transition matrix and measurement matrix of the state variable;

S503: Initialize the covariance matrix and set the moving target position of the first frame as the initial position and the initial speed as zero.

Based on the above technical solution, the present invention also provides a KVM system for monitoring and managing the flow of people in scenic spots. The system includes a processor, an FPGA chip, and a computer program stored on the FPGA chip and running on the processor. It is characterized in that , the HiSilicon chip model is XC7A100T-2FGG484I, and the following steps are implemented when executing the program:

S1: The three-frame difference method is used to detect the running target of the collected video image;

S2: Segment the running target of the video image after the running target detection has been completed;

S3: Take the tourist head area as the target feature, and set the enclosing rectangle of the second head area as the tracking window;

S4: Use Kalman filtering to predict the area of the next frame of the tracking window, and finally find the best matching object in the prediction area combined with the minimum Euclidean distance to achieve target tracking;

S5: When the next frame of image comes, if the tracked moving target prediction area is not within the shooting area, the count value will change, so as to count the running target.

The running target detection adopts the three-frame difference method, which includes the following steps:

S101: Use f _K-1 (x, y), f _K (x, y) and f _K+1 (x, y) to represent the k-1 th frame, the k th frame and the k+1 th frame image in (x , y) pixel grayscale of point;

S102: Differentiate f _K (x, y) and f _K-1 (x, y), make a difference between f _K+1 (x, y) and f _K (x, y), and record the difference image as D _K ( x,y), D _K+1 (x,y), D _K (x,y) are obtained by the following formulas:

S103: Perform AND operation on the difference images D _K (x, y) and D _K+1 (x, y) to obtain an image

S104: Use the maximum inter-class variance method to perform threshold segmentation to obtain a binary image

S105: Use the dilation operation to perform morphological processing, so that a clearer foreground moving target image Rn can be obtained.

Having adopted the above-mentioned technical scheme, the present invention has the following positive effects:

(1) the present invention adopts FPGA chip to carry out parallel operation to the video image, and the processing speed is fast and the delay is low, which ensures that the change of the flow of people in the scenic spot can be detected in real time.

(2) The camera is used to shoot vertically to solve the problem of crowded people and the mutual occlusion of moving objects. The moving object determination and tracking algorithm is based on the circular feature of the head of tourists, which improves the accuracy of moving object detection.

Description of drawings

In order to make the content of the present invention easier to understand clearly, the present invention will be described in further detail below according to specific embodiments and in conjunction with the accompanying drawings, wherein

Fig. 1 is the working schematic diagram of FPGA internal video image processing of the present invention;

Fig. 2 is a three-frame difference method process diagram of the present invention;

3 is a schematic diagram of the experiment of the three-frame difference method of the present invention

Fig. 4 is the moving target segmentation flow chart of the present invention;

5 is a schematic diagram of a moving target segmentation experiment of the present invention;

Fig. 6 is the flow chart of moving target tracking of the present invention;

7 is a schematic diagram of a moving target tracking experiment of the present invention;

FIG. 8 is a graph of tourist counts in a simulated scenic spot of the present invention.

Detailed ways

The above description is only an overview of the technical solutions of the present invention, in order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand , the following specific embodiments of the present invention are given.

The technical solutions of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. If there is no conflict, the embodiments of the present application and the technical features in the embodiments may be combined with each other.

The term "and/or" in this document is only an association relationship to describe associated objects, indicating that there can be three kinds of relationships, for example, A and/or B, which can mean that A exists alone, A and B exist at the same time, and A and B exist independently. B these three cases. In addition, the character "/" in this text generally indicates that the related objects are an "or" relationship.

(Example 1)

A KVM method for monitoring and managing the flow of people in scenic spots is applied to an electronic device, in the present embodiment, the specific FPGA chip of the electronic device.

S1: The three-frame difference method is used to detect the running target of the collected video image;

然后在对 差分图像D_K(x,y)和D_K+1(x,y)进行与操作，得到图像

其次采用最大类间 方差法进行阈值分割得到二值图像

最后用膨胀运算进行形态化处理， 这样能获得较为清晰的前景运动目标图像Rn。As shown in Figure 2, firstly use f _K-1 (x, y), f _K (x, y) and f _K+1 (x, y) to represent the k-1th frame, the kth frame and the k+1th frame The pixel grayscale of the frame image at (x, y) point, then f _K (x, y) and f _K-1 (x, y) are differentiated, f _K+1 (x, y) and f _K (x ,y) as the difference, the difference image is recorded as D _K (x,y), D _K+1 (x,y), for example, D _K (x,y) is obtained by the following formula:

Then perform AND operation on the difference image D _K (x, y) and D _K+1 (x, y) to get the image

Secondly, the maximum inter-class variance method is used to perform threshold segmentation to obtain a binary image.

Finally, the dilation operation is used for morphological processing, so that a clearer foreground moving target image Rn can be obtained.

Figure 3 is a schematic diagram of the experiment of the three-frame difference method. The k-1 frame, the k frame and the k+1 frame are extracted as the operation targets. After the three-frame difference algorithm, the moving target image is finally obtained, which is the subsequent motion. Target tracking and counting lay the groundwork.

S2: Segment the running target of the video image after the running target detection has been completed;

得出质心坐标，x和y分别代表质心的横坐标和纵坐标，质心 个数则代表了游客目标人数，最后将质心之间横坐标的均值作为标点，画 垂直分割界限。Figure 4 is a flowchart of moving object segmentation. If the extracted moving object regions are adhered to each other, segmentation processing is performed. If the moving object region is determined to be a single moving object, no segmentation is required. The segmentation steps are divided into the acquisition of the circumscribed rectangle of the connected domain, the segmentation boundary statistics of the adhering crowd, and the segmentation. Connected domain circumscribed rectangle: first draw a circumscribed rectangle for the outer contour of the human body, and scan the rectangle to obtain the four-corner coordinates of the rectangle as X1n, X2n, Y1n, Y2n, where n is the sequence number of the sticking motion target rectangle, through the formula D ₁ =|X _2N -X _1N |,D ₂ =|Y _2N -Y _1N |, get the length and width of the bounding rectangle outline of the moving target, set the length threshold of the bounding rectangle as A, the width as B, and the aspect ratio as C. If the ratio of the length, width and length of the rectangle is within the set threshold range, it is determined as a single moving target. If it is not within the range, it means that there is more than one moving target, and the bounding rectangle needs to be divided. The length of the circumscribed rectangle usually exceeds the threshold due to the occlusion between the front and rear, so the vertical segmentation method is selected. The segmentation point statistics based on the circumscribed rectangle of the connected domain: first, the area of each circumscribed rectangle is calculated as m ₀₀ and the two first-order moments are denoted as m ₁₀ and m ₀₁ respectively, and then the centroid formula is used to calculate as

The centroid coordinates are obtained, x and y represent the abscissa and ordinate of the centroid, respectively, and the number of centroids represents the target number of tourists. Finally, the average of the abscissas between the centroids is used as a punctuation point to draw the vertical division boundary.

Figure 5 is a schematic diagram of the moving target segmentation experiment. The left side of the figure is the left-right adhesion of two people and the left-right adhesion of three people, and the right side is the result of the moving target segmentation, which clearly shows the number of moving objects.

S3: Take the tourist head area as the target feature, and set the enclosing rectangle of the second head area as the tracking window;

S4: Use Kalman filtering to predict the area of the next frame of the tracking window, and finally find the best matching object in the prediction area combined with the minimum Euclidean distance to achieve target tracking;

Figure 6 shows the flow chart of moving target tracking. A tracking algorithm combining Kalman filter prediction and minimum Euclidean distance is used. The algorithm steps are divided into four steps: first, select the target feature parameters, secondly establish and initialize the system model, and perform target prediction again. If the predicted target is the same as the matching target, the target will be updated. If the predicted target exceeds the monitoring area, the count value will change. The actual operation steps are as follows:

得出窗 口中心点(X,Y)，最后假设外接矩形中某点像素为f(x,y)，头部区域像素总 数为N,则头部区域像素均值为

Select target feature parameters: Select target feature parameters: After the moving target is segmented, first use the Canny edge detection algorithm to extract the edge features of each rectangular box, and then use the Hough transform to detect the circle in the upper half of the rectangular box and use it as a motion The head of the target, then make the detected moving target head as the smallest circumscribed rectangle and scan the coordinates on the four corners of the circumscribed rectangle as X1n, X2n, Y1n, Y2n, where n is the moving target head serial number, through the formula D _X = |X _2N -X _1N |, D _Y = |Y _2N -Y _1N |, calculate the length and width of the target head window, and use the formula

Obtain the center point (X, Y) of the window, and finally assume that a certain point pixel in the circumscribed rectangle is f(x, y), and the total number of pixels in the head area is N, then the average pixel value of the head area is

System model establishment and initialization: Since the time difference between adjacent frames is very small, the speed v of the moving object changes very little, then it is determined that the moving object is moving at a uniform speed. First, select the center point of the window (X, Y), the pixel mean w and the speed v As a state variable, then calculate the transition matrix and measurement matrix of the state variable, and finally initialize the covariance matrix and set the first frame moving target position as the initial position and the initial speed as zero;

Target prediction and matching: Based on the center point of the current moving target, Kalman filtering is used to predict the position of the center point of the next frame of moving target. When the next frame of image comes, a pre-search area with a radius of R is set based on the predicted center point, and within this range Search within the minimum Euclidean distance to obtain the best matching object;

Target update: take the best matching object as the initial position, update the target information, and repeat the above operations until the end.

As shown in Figure 7 is a schematic diagram of the moving target tracking experiment, which are the kth frame, the k+3th frame, the k+6th frame and the k+12th frame image, using the Kalman filter prediction and the minimum Euclidean distance combined tracking algorithm It can effectively track moving targets.

S5: When the next frame of image comes, if the tracked moving target prediction area is not within the shooting area, the count value will change, so as to count the running target.

As shown in Figure 8 is the tourist count map of the simulated scenic spot. The camera shooting area is from the railings on both sides to the ticket gate, that is, the area marked by the elliptical dotted line in the figure. It is assumed that the tourists are standing in the undetected area A before entering the park. As soon as it enters the detection area B, it enters the counting range, and the video image processing algorithm starts to work. According to the moving target tracking algorithm, the first position of the moving target entering the detection area B is used as the initial position of the moving target, and the position detection and position update are carried out in real time. The target appears in area C, and the number of people entering the scenic spot changes.

(Example 2)

Based on the same inventive concept as a KVM method for monitoring and managing the flow of people in scenic spots in the foregoing embodiment, the present invention also provides a KVM system for monitoring and managing the flow of people in scenic spots.

Specifically, it includes processor and FPGA chip and a computer program that is stored on the FPGA chip and can run on the processor, and is characterized in that the HiSilicon chip model is XC7A100T-2FGG484I.

As shown in Figure 1, taking Scenic Entrance 1 as an example, three cameras respectively collect three-way video images and transmit them to the PFGA chip. First, the FIFO memory saves the received video signal, while local loop-out is carried out through the buffer to ensure the signal output quality, while under the control of the timing, each frame of image is converted into a grayscale image through the RGB-YCbCr module, and then through the RGB-YCbCr module. The storage controller stores it in a certain address of the DDR3 memory, and then reads the data through the storage controller, performs moving target detection, moving target segmentation, moving target tracking and moving target counting, and finally passes the processing results through the SPI interface controller. The SPI bus is transmitted to the sending HiSilicon module for compression encoding, and the following steps are implemented when the program is specifically executed:

S1: The three-frame difference method is used to detect the running target of the collected video image;

S2: Segment the running target of the video image after the running target detection has been completed;

S3: Take the tourist head area as the target feature, and set the enclosing rectangle of the second head area as the tracking window;

S4: Use Kalman filtering to predict the area of the next frame of the tracking window, and finally find the best matching object in the prediction area combined with the minimum Euclidean distance to achieve target tracking;

S5: When the next frame of image comes, if the tracked moving target prediction area is not within the shooting area, the count value will change, so as to count the running target.

The running target detection adopts the three-frame difference method, which includes the following steps:

S101: Use f _K-1 (x, y), f _K (x, y) and f _K+1 (x, y) to represent the k-1 th frame, the k th frame and the k+1 th frame image in (x , y) pixel grayscale of point;

S102: Differentiate f _K (x, y) and f _K-1 (x, y), make a difference between f _K+1 (x, y) and f _K (x, y), and record the difference image as D _K ( x,y), D _K+1 (x,y), D _K (x,y) are obtained by the following formulas:

S103: Perform AND operation on the difference images D _K (x, y) and D _K+1 (x, y) to obtain an image

S104: Use the maximum inter-class variance method to perform threshold segmentation to obtain a binary image

S105: Use the dilation operation to perform morphological processing, so that a clearer foreground moving target image Rn can be obtained.

The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in further detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (6)

1. a kind of KVM method that is used for scenic spot people flow monitoring and management, is applied to an electronic equipment, it is characterized in that: described method comprises S1: The three-frame difference method is used to detect the running target of the collected video image; S2: Segment the running target of the video image after the running target detection has been completed; S3: The head area of the tourist is used as the target feature, and the enclosing rectangle of the second head area is set as the tracking window; S4: Use Kalman filtering to predict the area of the tracking window of the next frame, and finally find the best matching object in the prediction area combined with the minimum Euclidean distance to achieve target tracking; S5: When the next frame of image comes, if the tracked moving target prediction area is not within the shooting area, the count value changes, so as to count the running target. The running target detection adopts the three-frame difference method, which includes the following steps: S101: Use f _K-1 (x, y), f _K (x, y) and f _K+1 (x, y) to represent the k-1 th frame, the k th frame and the k+1 th frame in (x , y) pixel grayscale of point; S102: Differentiate f _K (x, y) and f _K-1 (x, y), make a difference between f _K+1 (x, y) and f _K (x, y), and record the difference image as D _K ( x,y), D _K+1 (x,y), D _K (x,y) are obtained by the following formulas:

S103: Perform AND operation on the difference images D _K (x, y) and D _K+1 (x, y) to obtain an image

S104: Use the maximum inter-class variance method to perform threshold segmentation to obtain a binary image

S105: Perform morphological processing by using an expansion operation, so that a clearer foreground moving target image Rn can be obtained. 2. a kind of KVM method that is used for scenic spot people flow monitoring and management according to claim 1, the running target in described step S2 is divided, it is characterized in that if the moving target area of extraction is adhered to each other then carry out segmentation processing, if If it is determined to be a single moving target, no segmentation is required. The steps of target segmentation include: S201: Circumscribed rectangle of the connected domain: first, draw a circumscribed rectangle for the outer contour of the human body, and scan the rectangle to obtain the four-corner coordinates of the rectangle as X1n, X2n, Y1n, Y2n, where n is the sequence number of the sticking motion target rectangle; S202: By the following formula: D ₁ =|X _2N -X _1N |, D ₂ =|Y _2N -Y _1N | The length and width of the bounding rectangle outline of the moving target are obtained, and the length threshold of the bounding rectangle is set as A, the width as B, and the aspect ratio as . If the ratio of length, width and length to width of the circumscribed rectangle is within the set threshold range, it is determined as a single moving target. If it is not within the range, it means that there is more than one moving target, and the circumscribed rectangle needs to be divided. S203: adopt the vertical division method, calculate the area of each circumscribed rectangle and mark it as m ₀₀ and the two first-order moments are marked as m ₁₀ and m ₀₁ respectively; S204: Using the centroid formula to calculate as

The centroid coordinates are obtained, x and y represent the abscissa and ordinate of the centroid, respectively, and the number of centroids represents the target number of tourists. Finally, the average of the abscissas between the centroids is used as a punctuation point to draw the vertical division boundary. 3. a kind of KVM method for scenic spot people flow monitoring and management according to claim 1, the target tracking in described step S4 is characterized in that comprising: S301: Select target feature parameters; S302: System model establishment and initialization; S303: target prediction and matching: based on the center point of the current moving target, use Kalman filtering to predict the position of the center point of the next frame of moving target, when the next frame of image comes, set a pre-search area with a radius of R based on the predicted center point, Search with the minimum Euclidean distance within this range to obtain the best matching object; S304: target update: take the best matching object as the initial position, update the target information, repeat the above operation until the end. 4. a kind of KVM method for scenic spot people flow monitoring and management according to claim 3, the selection target characteristic parameter in described step S301 is characterized in that comprising: S401: First, use the Canny edge detection algorithm to extract the edge features of each rectangular frame; S402: secondly, use Hough transform to detect the circle in the upper half of the rectangular frame and use it as the head of the moving target; S403: Make the detected head of the moving target a minimum circumscribed rectangle and scan the coordinates on the four corners of the circumscribed rectangle as X1n, X2n, Y1n, Y2n, where n is the sequence number of the moving target head, and calculate the target head window by the following formula length and width: D ₁ =|X _2N -X _1N |, D ₂ =|Y _2N -Y _1N | and by the formula:

Get the window center point (X,Y). S404: Assuming that the pixel of a certain point in the circumscribed rectangle is f(x, y), and the total number of pixels in the head area is N, the average pixel value of the head area is

5. a kind of KVM method for scenic spot people flow monitoring and management according to claim 3, the system model establishment and initialization in described step S302 is characterized in that comprising: S501: The time difference between adjacent frames is very small, resulting in a small change in the speed v of the moving object, then it is determined that the moving object is moving at a uniform speed, and the center point of the window (X, Y), the pixel mean w and the speed v are selected as state variables; S502: Calculate the transition matrix and measurement matrix of the state variable; S503: Initialize the covariance matrix and set the first frame moving target position as the initial position and the initial speed as zero. 6. A KVM system for monitoring and managing the flow of people in scenic spots, comprising a processor and an FPGA chip and a computer program stored on the FPGA chip and running on the processor, characterized in that the Hisilicon chip model is XC7A100T-2FGG484I, implement the following steps when executing the program: S1: The three-frame difference method is used to detect the running target of the collected video image; S2: Segment the running target of the video image after the running target detection has been completed; S3: The head area of the tourist is used as the target feature, and the enclosing rectangle of the second head area is set as the tracking window; S4: Use Kalman filtering to predict the area of the tracking window of the next frame, and finally find the best matching object in the prediction area combined with the minimum Euclidean distance to achieve target tracking; S5: When the next frame of image comes, if the tracked moving target prediction area is not within the shooting area, the count value changes, so as to count the running target. The running target detection adopts the three-frame difference method, which includes the following steps: S101: Use f _K-1 (x, y), f _K (x, y) and f _K+1 (x, y) to represent the k-1 th frame, the k th frame and the k+1 th frame in (x , y) pixel grayscale of point; S102: Differentiate f _K (x, y) and f _K-1 (x, y), make a difference between f _K+1 (x, y) and f _K (x, y), and record the difference image as D _K ( x,y), D _K+1 (x,y), D _K (x,y) are obtained by the following formulas:

S103: Perform AND operation on the difference images D _K (x, y) and D _K+1 (x, y) to obtain an image

S104: Use the maximum inter-class variance method to perform threshold segmentation to obtain a binary image

S105: Perform morphological processing by using an expansion operation, so that a clearer foreground moving target image Rn can be obtained.

Images