CN101106700A - Device and method for capturing intelligent target details in video surveillance system - Google Patents

Abstract

The invention relates to an intelligent target details acquisition device of video monitoring systems and the method thereof. The device uses one or a plurality of fixed cameras to monitor the whole monitored area, and uses one movable camera to capture target details in the set region. First of all, the invention obtains moving targets in the region from the images captured by the fixed cameras through a processing module; then tracks the separated moving targets and determines the motion state; if any target enters the set region, then according to the preset position parameters and focal length parameters of the region, the invention will adjust the direction and the focal length of the movable camera to capture detailed images of the target in the region; at the same time, the invention can conduct storage and remote transmission of the images as required. The invention has the advantages that the invention provides the video monitoring device which not only can monitor a large area, but also can automatically capture details of physical targets, and the method thereof, solving the contradiction between the monitored scope and the monitored target details in the prior art.

Description

Device and method for capturing intelligent target details in video surveillance system

technical field

The invention belongs to the technical field of video monitoring, and relates to a device and method for capturing intelligent target details in a video monitoring system.

Background technique

At present, in the field of video surveillance technology, most video surveillance videos can only record the motion state of the target in the surveillance scene, but cannot provide clear target detail information. Like this, make the use value of video recording greatly reduce. At present, there are two main methods to improve the definition of target details: one is to use high-quality imaging equipment to improve the imaging resolution, or to provide image information of target details by setting multiple cameras in the scene. The disadvantage is that the cost of the video surveillance system is increased; another method is to use a video surveillance system with a PTZ camera, but at present this kind of method works through manual adjustment, and the disadvantage is that it cannot achieve intelligence. The efficiency of detail image acquisition is relatively low.

Therefore, it is necessary to solve the contradiction between the scope of video surveillance and the collection of specific target details in the existing surveillance system, and provide a video surveillance system that can effectively solve the contradiction between the two without increasing the cost.

Contents of the invention

The object of the present invention is to provide an intelligent target detail capture device and method in a video surveillance system, which can not only monitor and record large scenes, but also intelligently and efficiently capture target detail information entering a set area, solving the problem of The contradiction between large-scale scene monitoring and target detail information capture improves the use value of surveillance video.

In order to achieve the above object, technical scheme of the present invention is as follows:

The intelligent target detail capture device in the video surveillance system is used to capture the panoramic image of the monitoring area and the target detail image of the set area. The device consists of an imaging module 1, an input module 2, a processing module 3, a control module 4, and a storage module 5 and a transmission module 6; the imaging module 1 is used for imaging of the panorama and target details, and outputs video images of the panorama and target details after imaging; the input module 2 is connected between the imaging module 1 and the processing module 3 for capturing The video image that imaging module 1 outputs; Processing module 3 is connected between input module 2 and control module 4, is used for processing the panorama video image captured by input module 2, and detects whether target appears in setting area; Control module 4 Be connected between processing module 3 and imaging module 1, be used for controlling imaging module 1 to capture the image of target detail; Storage module 5 and transmission module 6 are respectively connected to input module 2, and storage module 5 is used for storing all captured video images, The transmission module 6 is used for transmitting video images to remote terminals.

The imaging module 1 is made up of one or more fixed cameras 7 and a pan-tilt camera 8, the fixed camera 7 is connected with the input module 2 for panoramic imaging, the pan-tilt camera 8 is connected with the input module 2 and the control module 4 connection for imaging of target details.

The method for capturing intelligent target details in a video surveillance system includes the following steps: a fixed camera 7 performs panoramic imaging of a monitoring area; an input module 2 captures a panoramic video image of the monitoring area, and outputs the image to a processing module 3; the processing module 3 Utilize moving average method to set up background model, carry out area numbering to the set area in monitoring area, and record pan-tilt camera 8 to capture the position parameter and focal length parameter of this corresponding number area; Processing module 3 utilizes background removal method to extract moving target; Processing Module 3 tracks and detects the moving object, and judges whether the moving object enters the setting area; if the moving object does not enter the setting area, then continue to track the moving object; if the moving object enters the setting area, then output the setting The number and control parameters of the area are given to the control module 4; the control module 4 controls the pan-tilt camera 8 to rotate to the position corresponding to the area recorded in advance by the processing module 3 according to the received number and control parameters of the set area, and controls its adjustment Its own focusing parameters; the pan-tilt camera 8 captures images of target details; the storage module 5 and the transmission module 6 store and transmit the captured images.

After the input module 2 captures the panoramic video image of the monitoring area and outputs the image to the processing module 3, a storage module 5 and a transmission module 6 store and transmit the captured image.

The beneficial effects of the present invention are: the fixed camera is used to capture the panoramic video image of the monitoring area, and the control module is used to control the pan-tilt camera to intelligently and efficiently capture the target detail video image of the set area, which solves the problem of panoramic monitoring and design of the monitoring area. The contradiction between the capture of target details in a certain area improves the use value of surveillance video.

Description of drawings

Fig. 1 is a schematic structural diagram of an intelligent target detail capturing device in a video surveillance system of the present invention.

In the figure: 1. imaging module, 2. input module, 3. processing module, 4. control module, 5. storage module, 6. transmission module, 7. fixed camera, 8. pan-tilt camera.

Detailed ways

The present invention is described in further detail below in conjunction with accompanying drawing:

As shown in Figure 1, the intelligent target details capture device in the video monitoring system of the present invention is made up of imaging module 1, input module 2, processing module 3, control module 4, storage module 5 and transmission module 6, and imaging module 1 is made up of One or more fixed cameras 7 and a pan-tilt camera 8 are formed. One or more fixed cameras 7 are responsible for the imaging of the panorama, and the pan-tilt camera 8 is responsible for the imaging of the target details. The input video images are captured by the input module 2, and the processing module 3 processes the images obtained by each fixed camera to detect each Whether there is a target in the setting area, the control module 4 controls the pan-tilt camera 8 to capture the target detail image, and the storage module 5 and the transmission module 6 control the storage and transmission of all collected images.

The realization of the intelligent target detail capture method in the video monitoring system of the present invention is as follows:

1. The imaging module 1 is composed of one or more fixed cameras 7 responsible for panoramic imaging, and the camera direction and focal length of the pan-tilt camera 8 can be adjusted, which is responsible for the imaging of target details.

2. The input module 2 is composed of a video capture card and a capture program, and is responsible for image input and image format conversion. Set the resolution, image format, sampling interval and other content of the collected image by calling the initialization function of the driver; call the collection function to complete the collection of all images.

3. The processing module 3 realizes the following functions:

1. Use the mouse to set each camera to capture the set area in the scene and number the area. At the same time, record the position parameters and focal length parameters of the corresponding area captured by the pan/tilt camera 8.

The background model is established by the moving average method, and the background changes due to changes in scene content or changes in lighting conditions are updated.

The motion area in the image captured by each camera is obtained by subtracting the background model from the current frame.

The processing method is as follows:

The moving target is extracted by the background removal method, and the processing steps are as follows:

1) Extraction of motion area: extract motion area by background removal method:

$\mathrm{<span\; class="notranslate">\; V</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&mu;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; else</span>}\end{array}\right.$

Among them, V(x, y, t) is the binary image of the change area, I(x, y, t) is the input image at time t, μ(x, y, t) is the background model at time t, and T is the threshold.

2) Background model update algorithm: The improved moving average method is used to update the background, and the processing method is as follows:

μ _t ＝Mμ _t-1 +(1-M)(αI _t +(1-α)μ _t-1 )

Among them, M is the motion region template, μ _t and μ _t-1 are the background models at time t and time t-1, and α is the update rate.

Detect and track the moving target, and output the number of the area to the control module when the target is found in the set area.

Video tracking processing, tracking the moving target and judging the motion state, the processing steps are as follows:

a) Calculation of eigenvalues, for the moving target extracted after the 1) operation in step 1, calculate its eigenvalues, including centroid and tracking window.

Select the centroid and tracking window size as feature values to track the object.

The first is to set the tracking window of the moving target, that is, use the circumscribed rectangle of the target as the tracking window.

L＝x _max -x _min

W = y _max - y _min

Among them, x _max and x _min are the maximum and minimum coordinates of the target in the horizontal direction, respectively, and y _max and y _min are the maximum and minimum coordinates of the target in the vertical direction, respectively.

After each tracking window is marked, the center of mass of the target in the window is calculated respectively, and the input image is f(x, y), as shown in the following formula:

f(x, y) is the image of the moving target in the tracking window, and the centroid of the window can be calculated. If the target pixel (x, y) ∈ S, the coordinates of the centroid of the window are

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; xf</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ,</span>$ $\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; yf</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}$

b) Use the Kalman filter to establish the motion model of the system and define the state vector to predict the possible position of the moving target in the next frame.

This processing step uses Kalman filtering to establish a motion estimation model for the system. Using Kalman filter for motion estimation can reduce noise interference, narrow the search range of feature extraction, and only need to detect the current tracking window, reducing the amount of calculation.

Assume that the state vector s k+1 at time k+ ₁ in the model is composed of the transfer function and noise of the vector s _k at time k. The observation vector is determined by the observation function and noise of the vector s k ₊ 1 at time k+1.

The state equation is as follows

s _k+1 = _Ask +w _k

measurement equation

z _k+1 ＝Cs _k+1 +v _k+1

In the formula, w _k and v _k+1 are normal white noise with zero mean.

s _k is the state vector consisting of an eight-dimensional vector:

${\mathrm{<span\; class="notranslate">\; the\; s</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\stackrel{<span\; class="notranslate">\; .</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\stackrel{<span\; class="notranslate">\; .</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}}_{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; xk</span>}}\\ {\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; yk</span>}}\\ {\stackrel{<span\; class="notranslate">\; .</span>}{\mathrm{<span\; class="notranslate">\; L</span>}}}_{\mathrm{<span\; class="notranslate">\; xk</span>}}\\ {\stackrel{<span\; class="notranslate">\; .</span>}{\mathrm{<span\; class="notranslate">\; L</span>}}}_{\mathrm{<span\; class="notranslate">\; yk</span>}}\end{array}\right]$

分别为质心坐标在x，y方向上的单位位移，L_xk，L_yk分别为跟踪窗口在x，y方向上的宽度，

分别为跟踪窗口宽度在x，y方向上的单位位移。In the formula, x _k , y _k are the coordinates of the center of mass of the target,

are the unit displacement of the centroid coordinates in the x and y directions respectively, L _xk and L _yk are the widths of the tracking window in the x and y directions respectively,

are the unit displacements of the tracking window width in the x and y directions, respectively.

z _k+1 is the observation vector, which is composed of four-dimensional vectors.

${\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; xk</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; yk</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\end{array}\right]$

Since the sampling interval is very short, it can be approximated that the moving speed of the moving target is constant, and the size of the tracking window does not change much, then the state transition matrix A is:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\begin{array}{cccccccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; t</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; t</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; t</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; t</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}$

The observation matrix C is:

$\mathrm{<span\; class="notranslate">\; C</span>}<span\; class="notranslate">\; =</span>\begin{array}{cccccccc}\mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}& \mathrm{<span\; class="notranslate">\; 0</span>}\end{array}$

c) Feature matching, define the similarity function of the target, use the change of the target between relative frames, use the feature value to calculate the value of the similarity function, and judge whether it is the same tracking target.

First define the center of mass of the i-th target in the kth frame and the centroid distance function of the j-th target in the k+1th frame:

$\mathrm{<span\; class="notranslate">\; D.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>}$

In the formula: $\left|c_k^i{c}_{k+1}^j\right|=\sqrt{{(x_k^i-x_{k+1}^j)}^2+{({\mathrm{the\; y}}_k^i-{\mathrm{the\; y}}_{k+1}^j)}^2}$ The area difference function, that is, the window area of the i-th target in the k-th frame is compared with the window area of the j-th target in the k+1th frame:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>}$

In the formula: $|a_k^i-a_{k+1}^j|=|L_{\mathrm{xk}}^i\&times;L_{\mathrm{yk}}^i-L_{\mathrm{xk}+1}^j\&times;L_{\mathrm{yk}+1}^j|$

Define the similarity function

Δ(i,j)=γD(i,j)+ξA(i,j)

In the formula, γ and ξ are weights, and satisfy γ>ξ, γ+ξ=1, Δ(i, j)≤1.

If D(i, j) is smaller, the target is closer, and A(i, j) is smaller, the target shape is closer, and Δ(i, j) is smaller, indicating that the two targets are similar maximum. Set the threshold T _Δ of the similarity function as the basis for whether they are the same target.

d) Model update, updating the motion model as the input of the Kalman filter of the next motion model.

When the minimum value of the similarity function is found, the follow-up of the same target has been found, that is, the j-th target in the k+1th frame can be considered as the i-th target in the k-th frame, that is, the two are the same target. At this time, use the feature value of the jth target in the k+1th frame as the input of the motion model to estimate the next frame, and so on to complete the tracking of the model.

2) When the moving target is detected to enter the setting area, the number and control parameters of the setting area are output.

Four, the processing steps of control module 4 are as follows:

According to the area number generated by the image processing module, the corresponding position parameters and focal length parameters are searched, and the position parameters and focal length parameters are output to the control module of the pan-tilt camera through the control port.

Its processing steps are as follows:

1. Search for the corresponding position parameter and focal length parameter according to the area number generated by the image processing module, and output the position parameter and focal length parameter to the control module of the pan/tilt camera through the control port.

2. Through the communication port, such as RS-232 or USB, output control commands to control the pan/tilt camera to rotate to the set position and adjust the camera focus parameters.

Five, the storage module 5 compresses the images collected by each fixed camera 7 and the pan-tilt camera 8 and stores them in the hard disk. The collected images can be compressed and stored in formats such as MPEG-1, MPEG-4, WAVELET, and H.263

Sixth, the transmission module 6 can transmit the images collected by each camera to the remote terminal through the network through Ethernet according to actual needs.

Claims (4)

1. An intelligent target detail capturing device in a video monitoring system is used for capturing a panoramic image of a monitored area and a target detail image of a set area, and is characterized by comprising an imaging module (1), an input module (2), a processing module (3), a control module (4), a storage module (5) and a transmission module (6); the imaging module (1) is used for imaging the panorama and the target details and outputting video images of the imaged panorama and the target details; the input module (2) is connected between the imaging module (1) and the processing module (3) and is used for capturing a video image output by the imaging module (1); the processing module (3) is connected between the input module (2) and the control module (4) and is used for processing the panoramic video image captured by the input module (2) and detecting whether a target appears in a set area; the control module (4) is connected between the processing module (3) and the imaging module (1) and is used for controlling the imaging module (1) to capture an image of the target details; the storage module (5) and the transmission module (6) are respectively connected to the input module (2), the storage module (5) is used for storing all captured video images, and the transmission module (6) is used for transmitting the video images to a remote terminal.

2. The intelligent target detail capturing device in the video monitoring system according to claim 1, wherein the imaging module (1) is composed of one or more fixed cameras (7) and a pan-tilt camera (8), the fixed cameras (7) are connected with the input module (2) for panoramic imaging, and the pan-tilt camera (8) is connected with the input module (2) and the control module (4) for target detail imaging.

3. An intelligent target detail capturing method in a video monitoring system is characterized by comprising the following steps:

a fixed camera (7) images the panorama of the monitored area;

the input module (2) captures a panoramic video image of a monitored area and outputs the image to the processing module (3);

the processing module (3) establishes a background model by using a moving average method, performs area numbering on a set area in a monitoring area, and records a position parameter and a focal length parameter of a corresponding numbered area captured by a holder camera (8);

the processing module (3) extracts a moving target by using a background removal method;

the processing module (3) tracks and detects the moving target and judges whether the moving target enters a set area or not;

if the moving target does not enter the set area, continuing to track the moving target;

if the moving target enters a set area, outputting the number and the control parameters of the set area to a control module (4);

the control module (4) controls the holder camera (8) to rotate to the position, corresponding to the area, recorded in advance by the processing module (3) according to the received serial number and the control parameter of the set area, and controls the holder camera to adjust the focusing parameter of the holder camera;

a pan-tilt camera (8) captures an image of the target details;

the storage module (5) and the transmission module (6) store and transmit the captured images.

4. An intelligent target detail capturing method in a video surveillance system according to claim 3, characterized in that after the step of capturing panoramic video images of the monitored area by the input module (2) and outputting the images to the processing module (3), the method further comprises the step of storing and transmitting the captured images by the storage module (5) and the transmission module (6).

Images