A security inspection monitoring image enhancement method and system based on machine vision

Abstract

The present application relates to the technical field of image processing, and more particularly to a security monitoring image enhancement method and system based on machine vision. The method comprises the following steps: collecting each frame of security monitoring image, and obtaining the dynamic region of each frame of security monitoring image; obtaining the behavior motion complexity of the dynamic region according to the area change and the number of corner points of the dynamic region; performing clustering analysis on the corner points of the dynamic region, and obtaining the behavior confusion of the dynamic region; obtaining the behavior complexity degree of the dynamic region based on the behavior motion complexity and the behavior confusion; obtaining the blur degree of the dynamic region according to the behavior complexity degree; obtaining the enhancement coefficient of the current frame of security monitoring image according to the blur degree of all dynamic regions of the current frame of security monitoring image, enhancing the current frame of security monitoring image according to the enhancement coefficient, and obtaining the current frame of security monitoring enhanced image. After the image is enhanced, the action behavior of the personnel can be more easily recognized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and in particular to a security monitoring image enhancement method and system based on machine vision. BACKGROUND

[0002] With the rapid development of society and the continuous progress of technology, the application of security monitoring systems in the field of public safety has also been significantly expanded, especially in transportation hubs such as airports, stations, and subway stations. These places are usually the focus of public safety management and emergency response, so efficient, real-time monitoring image processing and analysis systems have become indispensable tools to ensure public safety. In these environments, security monitoring systems not only need to deal with a large amount of monitoring data of the crowd, but also need to handle possible emergencies or potential security threats. For example, in stations or subway stations, the monitoring system needs to quickly respond to whether the behavior of passengers is abnormal to prevent stampede accidents or other emergencies, so in order to ensure safety in these complex environments, traditional security monitoring systems can no longer meet the growing demand, especially when faced with a large amount of complex real-time image data, effective real-time image processing and analysis can not only provide detailed dynamic pictures, but also quickly identify abnormal behavior and potential threats.

[0003] In the security monitoring system, due to the frequent and complex activities of personnel in the monitoring area, and when the personnel move quickly, the camera cannot capture a clear image in a short time, which easily leads to motion blur in the security monitoring video. This blur phenomenon causes important dynamic information in the security monitoring image to be lost, making it difficult to identify the action behavior of personnel in the monitoring area and unable to accurately extract potential threats. SUMMARY

[0004] In order to solve the technical problem that the monitoring area is complex due to personnel and when the personnel move quickly, it is difficult to capture a clear image in a short time, which easily leads to motion blur in the image, making it difficult to identify the action behavior of personnel, the present application provides a security monitoring image enhancement method and system based on machine vision.

[0005] In a first aspect, the present application provides a security monitoring image enhancement method based on machine vision, which adopts the following technical solution:

[0006] A security monitoring image enhancement method based on machine vision, comprising the steps of:

[0007] Collecting each frame of security monitoring image; obtaining each dynamic area of each frame of security monitoring image; obtaining the behavior motion complexity of each dynamic area in the current frame of security monitoring image according to the area change and the number of corner points of the dynamic area; obtaining the feature vector of each corner point of each dynamic area in the current frame of security monitoring image; clustering each dynamic area corner point in the current frame of security monitoring image based on the feature vector of the dynamic area, to obtain a plurality of clustering clusters; obtaining the behavior confusion of each dynamic area of the current frame of security monitoring image according to the number and distribution of the clustering clusters; obtaining the behavior complexity degree of each dynamic area of the current frame of security monitoring image by multiplying the behavior motion complexity and the behavior confusion.

[0008] According to the behavior complexity degree, obtaining the blur degree of each dynamic area of the current frame of security monitoring image; obtaining the enhancement coefficient of the current frame of security monitoring image according to the blur degree, and enhancing the current frame of security monitoring image to obtain the current frame of security monitoring enhancement image.

[0009] The innovation of the present application is that each dynamic area of each frame of security monitoring image is obtained first, and the dynamic area is extracted, which helps to screen out the part with activity from the static image for subsequent analysis; the behavior motion complexity is obtained according to the area change and the number of corner points of the dynamic area; the behavior confusion of the dynamic area is obtained according to the number of behavior mode types of the dynamic area; then the product of the behavior motion complexity and the confusion is considered comprehensively, which helps to comprehensively evaluate the behavior complexity degree of the dynamic area, and can accurately evaluate the blur degree in the dynamic area; further, the enhancement coefficient of the dynamic area is obtained according to the blur degree of the dynamic area, and the image is deblurred, so that the security personnel can clearly see important details, and more quickly and efficiently identify abnormal behavior, and respond to emergencies in time.

[0010] Preferably, the obtaining of each dynamic area of each frame of security monitoring image comprises:

[0011] A preset difference threshold T is used to record the absolute value of the gray difference of any pixel point in the current frame of security monitoring image and the pixel point at the same position in the previous frame of security monitoring image as a first difference, if the ratio of the first difference to 255 is greater than or equal to the difference threshold T, the pixel point in the current frame of security monitoring image is recorded as a dynamic pixel point; the region composed of all adjacent dynamic pixel points in the current frame of security monitoring image is recorded as a dynamic region, to obtain each dynamic area of the current frame of security monitoring image.

[0012] The extraction of the dynamic area helps to screen out the part with activity from the static image, reduces the interference of irrelevant areas, and is used for subsequent analysis.

[0013] Preferably, the behavior motion complexity of each dynamic region in the current frame security monitoring image is obtained by:

[0014] , representing the behavior motion complexity of the i-th dynamic region in the current frame security monitoring image; representing the area of the i-th dynamic region in the current frame security monitoring image; representing the area of the i-th dynamic region in the current frame security monitoring image in the corresponding dynamic region in the previous frame security monitoring image; representing the number of corner points of the i-th dynamic region in the current frame security monitoring image; norm() represents a normalization function.

[0015] By monitoring the area change and the number of corner points of the dynamic region, the complexity of the behavior motion can be captured, which can help to identify the blur of the dynamic region.

[0016] Preferably, the feature vector of each corner point of each dynamic region in the current frame security monitoring image is obtained by:

[0017] obtaining the matching corner point of each corner point of each dynamic region in the current frame security monitoring image;

[0018] the Euclidean distance between the position coordinates of the j-th corner point of the i-th dynamic region in the current frame security monitoring image and the position coordinates of the matching corner point thereof is recorded as the moving distance of the j-th corner point; the included angle between the direction of the line connecting the position coordinates of the j-th corner point of the i-th dynamic region in the current frame security monitoring image and the position coordinates of the matching corner point thereof and the horizontal direction is recorded as the moving direction of the j-th corner point; the moving distance, the moving direction and the coordinate position of the j-th corner point of the i-th dynamic region in the current frame security monitoring image are taken as the feature vector of the j-th corner point.

[0019] Preferably, the matching corner point of each corner point of each dynamic region in the current frame security monitoring image is obtained by:

[0020] SIFT descriptors are generated for the corner points of each dynamic region of each frame security monitoring image, and then a feature matching algorithm is used to match the corner points with similar SIFT descriptors in the corresponding dynamic region of the previous frame security monitoring image to obtain the matching corner point of each corner point of each dynamic region in the current frame security monitoring image.

[0021] Preferably, the behavior confusion of each dynamic region in the current frame security monitoring image is obtained by:

[0022] obtaining the adjacent clustering cluster of each clustering cluster of each dynamic region in the current frame security monitoring image;

[0023] , behavioral complexity of the i-th dynamic region of the current frame of the security inspection monitoring image; Euclidean distance between the n-th cluster center of the i-th dynamic region of the current frame of the security inspection monitoring image and the m-th neighboring cluster center thereof; number of neighboring cluster centers of the n-th cluster center of the i-th dynamic region of the current frame of the security inspection monitoring image; number of cluster centers of the i-th dynamic region of the current frame of the security inspection monitoring image; exp() represents an exponential function with a natural constant as a base number; norm() represents a normalization function.

[0024] The behavioral complexity reflects the number of motion types of the personnel in the monitoring region. If the complexity is high, it may mean that there are a large number of behavior patterns in the region, which can reflect that the region is more likely to be blurred.

[0025] Preferably, the method further comprises:

[0026] obtaining a sequence of behavioral complexity of the i-th dynamic region of the current frame of the security inspection monitoring image;

[0027] , blurriness of the i-th dynamic region of the current frame of the security inspection monitoring image; behavioral complexity of the i-th dynamic region of the current frame of the security inspection monitoring image; number of data in the sequence of behavioral complexity of the i-th dynamic region of the current frame of the security inspection monitoring image; d+1-th data value in the sequence of behavioral complexity of the i-th dynamic region of the current frame of the security inspection monitoring image; d-th data value in the sequence of behavioral complexity of the i-th dynamic region of the current frame of the security inspection monitoring image; norm() represents a normalization function.

[0028] According to the behavioral complexity, the blurriness of the image is derived, so that it can be judged whether the image clarity needs to be enhanced, and the difficulty of behavior recognition caused by unclear image can be effectively avoided.

[0029] Preferably, the method further comprises:

[0030] A preset number of historical frame images G is set; a matching relationship corresponding to a dynamic region between a current frame security monitoring image and G previous frame security monitoring images is obtained according to an optical flow method, so that a corresponding dynamic region of each dynamic region of the current frame security monitoring image in the G previous frame security monitoring images is obtained; and the behavior complexity of the i th dynamic region of the current frame security monitoring image and the G previous frame security monitoring images is arranged according to an image acquisition sequence, so that a behavior complexity sequence of the i th dynamic region of the current frame security monitoring image is obtained.

[0031] Preferably, the method further comprises: obtaining an enhancement coefficient of the current frame security monitoring image; and enhancing the current frame security monitoring image according to the enhancement coefficient to obtain a current frame security monitoring enhanced image.

[0032] The enhancement coefficient of the current frame security monitoring image is represented by K; The blur degree of the i th dynamic region of the current frame security monitoring image is represented by B i; The average of the blur degrees of all dynamic regions of the current frame security monitoring image is represented by B; The number of dynamic regions of the current frame security monitoring image is represented by N; the enhancement coefficient of the current frame security monitoring image is set as the sharpening intensity of each pixel point in all dynamic regions of the current frame security monitoring image, and the sharpening intensity of each pixel point in the non-dynamic region of the current frame security monitoring image is set as 0; and the current frame security monitoring image is enhanced by using a Laplacian sharpening algorithm based on the sharpening intensity of each pixel point in the current frame security monitoring image to obtain the current frame security monitoring enhanced image.

[0033] In a second aspect, the present application provides a security monitoring image enhancement system based on machine vision, which adopts the following technical scheme:

[0034] The security monitoring image enhancement system based on machine vision comprises a processor and a memory, and the memory stores computer program instructions; when the computer program instructions are executed by the processor, the above-mentioned security monitoring image enhancement method based on machine vision is realized.

[0035] By adopting the above-mentioned technical scheme, the above-mentioned security monitoring image enhancement method based on machine vision is generated into a computer program and stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and the use is facilitated.

[0036] ​The application has the following technical effects: the application first acquires each dynamic region of each frame of security monitoring image, and acquires behavior motion complexity according to the area change of the dynamic region and the number of corner points; acquires behavior confusion of the dynamic region according to the number of personnel motion behavior types of the dynamic region; then comprehensively considers the product of the behavior motion complexity and the confusion, which is helpful to comprehensively evaluate the behavior complexity degree of the dynamic region and can accurately evaluate the blur degree in the dynamic region; further, according to the blur degree of the dynamic region, an enhancement coefficient of the dynamic region is acquired and the image is deblurred, so that the security personnel can clearly see important details and more quickly and efficiently identify abnormal behaviors and timely respond to emergencies. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is a method flowchart in a security monitoring image enhancement method based on machine vision. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are some of the embodiments of the application, but not all the embodiments of the application.

[0039] The embodiments of the application disclose a security monitoring image enhancement method based on machine vision, referring to Figure 1 , including steps S1-S4:

[0040] S1: acquiring each frame of security monitoring image.

[0041] In the embodiments of the application, a camera is installed at a security checkpoint to acquire each frame of image in the security process, which is recorded as each frame of security monitoring image.

[0042] S2: acquiring each dynamic region of each frame of security monitoring image; acquiring behavior motion complexity of each dynamic region in the current frame of security monitoring image according to the area change of the dynamic region and the number of corner points; clustering the corner points of the dynamic region to obtain a plurality of clustering clusters; acquiring behavior confusion of each dynamic region of the current frame of security monitoring image according to the number and distribution of the clustering clusters; and acquiring behavior complexity degree of each dynamic region of the current frame of security monitoring image based on the behavior motion complexity and the behavior confusion.

[0043] It should be noted that in the monitoring area, due to the complexity of personnel and when the personnel move quickly, the camera cannot capture a clear image in a short time, which is easy to cause motion blur in the security monitoring video image, thereby causing difficulty in recognizing the action behavior of the personnel in the monitoring area, therefore, the application first identifies the dynamic area in the security monitoring image, analyzes the personnel motion condition of the dynamic area, obtains the behavior complexity degree of the dynamic area, then according to the behavior complexity degree, the blur degree of the dynamic area is evaluated, and then the dynamic areas of different blur degrees are enhanced to different degrees, so that the action behavior of the personnel in the dynamic area can be recognized.

[0044] It should be further pointed out that whether the pixel point changes is judged by comparing the gray difference value of the pixel point in the current frame image and the pixel point at the same position in the previous frame image, the gray difference value is larger, which indicates that the pixel position changes due to the motion of the personnel, therefore, by comparing the pixel difference between adjacent frame images, the dynamic area can be accurately extracted, and the basic data for subsequent behavior analysis and motion complexity evaluation is provided.

[0045] In the embodiment of the application, each dynamic area of each security monitoring image is obtained:

[0046] The preset difference threshold T is 0.7, in other embodiments, the value of the difference threshold can be preset according to the specific implementation; the absolute value of the gray difference value of any pixel point in the current frame security monitoring image and the pixel point at the same position in the previous frame security monitoring image is recorded as the first difference, if the ratio of the first difference to 255 is greater than or equal to the difference threshold T, the pixel point in the current frame security monitoring image is recorded as a dynamic pixel point; the region composed of all adjacent dynamic pixel points in the current frame security monitoring image is recorded as a dynamic area, and each dynamic area of the current frame security monitoring image is obtained.

[0047] Similarly, each dynamic area of each security monitoring image is obtained.

[0048] It should be noted that complex actions such as personnel gathering, moving and interacting occur in the dynamic area, so that the greater the behavior motion complexity of the dynamic area, the more likely the dynamic area appears in a fuzzy state, and therefore, the corresponding relationship of the dynamic area between adjacent frame security monitoring images is obtained by using the optical flow method, if the area of any dynamic area in the current frame security monitoring image changes more compared with the corresponding dynamic area in the previous frame security monitoring image, it is indicated that complex actions such as personnel gathering, moving and interacting occur in the dynamic area; in addition, the corner point of the dynamic area represents the body contour edge and joint (such as elbow, knee, ankle and the like) of the personnel in the dynamic area, so if the number of corner points of any dynamic area in the current frame security monitoring image is more, it is indicated that complex actions such as personnel gathering, deformation and interaction occur in the dynamic area; based on the above features, the behavior motion complexity of each dynamic area in the current frame security monitoring image is obtained.

[0049] In the embodiment of the application, the matching relationship of the dynamic area corresponding between the current frame security monitoring image and the previous frame security monitoring image is obtained according to the optical flow method, and the corresponding dynamic area of each dynamic area of the current frame security monitoring image in the previous frame security monitoring image is obtained.

[0050] The Harris corner point detection algorithm is used to detect the corner points in each dynamic area in each frame security monitoring image, and the corner points of each dynamic area of each frame security monitoring image are obtained.

[0051] The behavior motion complexity of each dynamic area in the current frame security monitoring image is obtained:

[0052] ;

[0053] In the formula, The behavior motion complexity of the i th dynamic area in the current frame security monitoring image is represented by S i ; The area of the i th dynamic area in the current frame security monitoring image is represented by A i ; The area of the corresponding dynamic area of the i th dynamic area in the current frame security monitoring image in the previous frame security monitoring image is represented by A i-1 ; The number of corner points of the i th dynamic area in the current frame security monitoring image is represented by C i ; and norm() represents a normalization function.

[0054] If the area of the dynamic area in the adjacent frame image changes greatly, it means that complex actions such as personnel gathering, moving and interacting may occur in the dynamic area, so the greater the behavior motion complexity of the dynamic area, the more behaviors need to be recognized, and the difficulty of behavior recognition increases; the corner point represents the body contour edge and joint (such as elbow, knee, ankle and the like) of the personnel, so if the number of corner points of the i th dynamic area in the current frame security monitoring image is more, it is indicated that complex actions such as personnel gathering, deformation and interaction occur in the dynamic area.

[0055] It should be noted that the corner points in the dynamic region represent some key feature points, such as the joints or contours of a person, the corner points of the dynamic region of the current frame image are matched with the corner points of the corresponding dynamic region in the previous frame image, and the Euclidean distance between each corner point of each dynamic region of the current frame security monitoring image and the matching corner point is calculated, so that the movement of the corner point can be quantified, and the movement distance of the person in the image can be directly reflected; then by obtaining the movement direction of each corner point of each dynamic region of the current frame security monitoring image, the movement direction of the corner point can be directly reflected; therefore, the movement distance, the movement direction and the position coordinates of each corner point in the dynamic region are taken as the feature vector of each corner point, and different feature vectors represent different movement modes;

[0056] Therefore, the feature vectors of the corner points in each dynamic region are used to cluster the corner points to form a plurality of clusters, and different clusters represent different movement behaviors in the dynamic region; therefore, the more the number of clustering clusters of the dynamic region and the more concentrated the distribution of all the clustering clusters of the dynamic region, the more diverse the movement behavior types in the dynamic region, and the more concentrated the distribution of different movement behaviors, and the more chaotic the movement behavior of the dynamic region.

[0057] In the embodiment of the application, the SIFT algorithm (Scale-Invariant Feature Transform) is used to generate a descriptor for each corner point of each dynamic region of each frame of security monitoring image, and then a feature matching algorithm is used to match the corner points with similar descriptors in the corresponding dynamic region of the previous frame of security monitoring image with the corner points of each dynamic region of the current frame of security monitoring image, so as to obtain the matching corner points of each corner point of each dynamic region of the current frame of security monitoring image; it should be noted that the corner points that cannot be matched are not analyzed subsequently.

[0058] The Euclidean distance between the position coordinates of the jth corner point of the ith dynamic region of the current frame of security monitoring image and the position coordinates of the matching corner point is recorded as the movement distance of the jth corner point of the ith dynamic region of the current frame of security monitoring image; the angle between the direction of the line connecting the position coordinates of the jth corner point of the ith dynamic region of the current frame of security monitoring image and the position coordinates of the matching corner point and the horizontal direction is recorded as the movement direction of the jth corner point of the ith dynamic region of the current frame of security monitoring image;

[0059] The movement distance, the movement direction and the coordinate position of each corner point of each dynamic region of the current frame of security monitoring image are taken as the feature vector, and all the corner points of each dynamic region of the current frame of security monitoring image are clustered to obtain a plurality of clustering clusters of each dynamic region of the current frame of security monitoring image;

[0060] For any one cluster of the i-th dynamic region of the current frame security monitoring image, the Euclidean distance between the cluster and all other clusters is obtained, and the four clusters with the closest distances are recorded as the neighboring clusters of the cluster;

[0061] The behavior confusion of each dynamic region of the current frame security monitoring image is obtained:

[0062]

[0063] In the formula, represents the behavior confusion of the i-th dynamic region of the current frame security monitoring image; represents the Euclidean distance between the n-th cluster of the i-th dynamic region of the current frame security monitoring image and the m-th neighboring cluster of the n-th cluster; represents the number of neighboring clusters of the n-th cluster of the i-th dynamic region of the current frame security monitoring image; represents the number of clusters of the i-th dynamic region of the current frame security monitoring image; exp() represents an exponential function with a natural constant as the base number; norm() represents a normalization function;

[0064] The greater the value of the behavior confusion of the i-th dynamic region of the current frame security monitoring image, the more the motion behavior patterns of the personnel in the dynamic region, and the greater the behavior confusion of the dynamic region;

[0065] The greater the value of the behavior confusion of the i-th dynamic region of the current frame security monitoring image, the closer the distance between each cluster of the dynamic region and its neighboring cluster, and the more concentrated the distribution of all the clusters of the dynamic region, that is, the more concentrated the distribution of different motion behaviors, and the greater the behavior confusion of the i-th dynamic region of the current frame security monitoring image.

[0066] It should be noted that the greater the behavior motion complexity and the greater the behavior confusion of each dynamic region of the current frame security monitoring image, the greater the behavior complexity of the dynamic region.

[0067] The product of the behavior confusion of the i-th dynamic region of the current frame security monitoring image and the behavior motion complexity of the i-th dynamic region of the current frame security monitoring image is taken as the behavior complexity of the i-th dynamic region of the current frame security monitoring image.

[0068] S3: According to the behavior complexity of each dynamic region of the current frame security monitoring image, the blur degree of each dynamic region of the current frame security monitoring image is obtained; and according to the blur degree of each dynamic region of the current frame security monitoring image, an enhancement coefficient of the current frame security monitoring image is obtained.

[0069] ​It should be noted that the greater the behavior complexity of any dynamic region of the current frame security monitoring image, the more likely the dynamic region of the current frame security monitoring image is blurred, and when the behavior complexity of the dynamic region continuously increases in the continuous frame security monitoring image, the more frequent and the more types of behaviors of the personnel at the dynamic region are, and thus the greater the blur degree of the dynamic region of the current frame security monitoring image, and thus the blur degree of each dynamic region of the current frame security monitoring image is obtained according to the above characteristics.

[0070] In the embodiment of the application, the preset number of historical frame images G=5, and in other embodiments, the value of G can be preset by the implementer according to the specific implementation; the matching relationship of the dynamic region between the current frame security monitoring image and the previous G frame security monitoring images is obtained according to the optical flow method, and the corresponding dynamic region of each dynamic region of the current frame security monitoring image in the previous G frame security monitoring images is obtained; the behavior complexity of the i th dynamic region of the current frame security monitoring image and the previous G frame security monitoring images is arranged according to the image acquisition sequence, and the behavior complexity sequence of the i th dynamic region of the current frame security monitoring image is obtained.

[0071] The blur degree of each dynamic region of the current frame security monitoring image is obtained.

[0072]

[0073] In the formula, represents the blur degree of the i th dynamic region of the current frame security monitoring image; represents the behavior complexity of the i th dynamic region of the current frame security monitoring image; represents the number of data in the behavior complexity sequence of the i th dynamic region of the current frame security monitoring image; represents the (d+1) th data value in the behavior complexity sequence of the i th dynamic region of the current frame security monitoring image; represents the d th data value in the behavior complexity sequence of the i th dynamic region of the current frame security monitoring image; norm() represents a normalization function;

[0074] If the behavior complexity of the i th dynamic region of the current frame security monitoring image is greater, the dynamic region is more likely to appear blurred, and at this time, the blur degree of the dynamic region is greater; represents the increase of the behavior complexity of the i th dynamic region in the continuous frame security monitoring image, and the greater the value, the more frequent and the more types of behaviors of the personnel at the i th dynamic region are, and thus the greater the blur degree of the i th dynamic region of the current frame security monitoring image is.​

[0075] It should be noted that the greater the blur degree of each dynamic region of the current frame security monitoring image, the greater the enhancement coefficient of the current frame security monitoring image, and therefore the enhancement coefficient of the current frame security monitoring image is obtained according to the blur degrees of all dynamic regions of the current frame security monitoring image.

[0076] In the embodiment of the present application, the enhancement coefficient of the current frame security monitoring image is obtained as follows:

[0077]

[0078] In the formula, represents the enhancement coefficient of the current frame security monitoring image; represents the blur degree of the i-th dynamic region of the current frame security monitoring image; represents the average of the blur degrees of all dynamic regions of the current frame security monitoring image; represents the number of dynamic regions of the current frame security monitoring image.

[0079] S4: The current frame security monitoring image is enhanced according to the enhancement coefficient of the current frame security monitoring image to obtain a current frame security monitoring enhanced image.

[0080] It should be noted that the current frame security monitoring image is processed using an adaptive Laplacian sharpening algorithm to enhance the details and edges of the blurred image; wherein the adaptive Laplacian sharpening algorithm gives each pixel point in the image a sharpening intensity, the value range of the sharpening intensity is between 0 and 1, the value of the sharpening intensity determines the mixing degree of the original image and the sharpened image, when the sharpening intensity is 0, there is no enhancement effect, the output image is the same as the original image, when the sharpening intensity is 1, the output image is completely determined by the sharpened image, and the enhancement effect is the strongest, therefore, different values between 0 and 1 represent different sharpening effects; and since the closer the enhancement coefficient of the current frame security monitoring image is to 1, the more blurred the current frame security monitoring image is, and the more it needs to be enhanced, therefore when the blur degree of the current frame security monitoring image is strong, the enhancement coefficient is close to 1, and more sharpening processing can be provided for the current frame security image.

[0081] In the embodiment of the present application, taking the current frame security monitoring image as an example, the sharpening intensity of each pixel point in the non-dynamic region of the current frame security monitoring image is 0, the enhancement coefficient of the current frame security monitoring image is taken as the sharpening intensity of each pixel point in all dynamic regions of the current frame security monitoring image, and the sharpening intensity of each pixel point in the current frame security monitoring image is obtained; based on the sharpening intensity of each pixel point in the current frame security monitoring image, the current frame security monitoring image is enhanced using the Laplacian sharpening algorithm to obtain a current frame security monitoring enhanced image.​

[0082] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, and thus: any equivalent changes made according to the structure, shape, principle of the present application should be encompassed within the protection scope of the present application.

Claims

1. A machine vision-based security monitoring image enhancement method, characterized in that, The method comprises the following steps: collecting each frame of security monitoring image; obtaining each dynamic region of each frame of security monitoring image; obtaining the behavior motion complexity of each dynamic region in the current frame of security monitoring image according to the area change of the dynamic region and the number of corner points, comprising: , behavioral motion complexity of the i-th dynamic region in the current frame of the security monitoring image; area of the i-th dynamic region in the current frame of the security monitoring image; area of the i-th dynamic region in the current frame of the security monitoring image corresponding to the dynamic region in the previous frame of the security monitoring image; number of corner points of the i-th dynamic region in the current frame of the security monitoring image; norm() represents a normalization function; obtaining the feature vector of each corner point of each dynamic region in the current frame of security monitoring image; clustering the corner points of each dynamic region in the current frame of security monitoring image based on the feature vector to obtain a plurality of clustering clusters; obtaining the behavior confusion of each dynamic region in the current frame of security monitoring image according to the number and distribution of the clustering clusters, comprising: obtaining the adjacent clustering cluster of each clustering cluster of each dynamic region in the current frame of security monitoring image; , behavioral turbulence of the i-th dynamic region representing the current frame security monitoring image; Euclidean distance between the n-th cluster center of the i-th dynamic region representing the current frame security monitoring image and the m-th neighboring cluster center thereof; number of neighboring cluster centers of the n-th cluster center of the i-th dynamic region representing the current frame security monitoring image; number of cluster centers of the i-th dynamic region representing the current frame security monitoring image; exp() represents an exponential function with a natural constant as a base number; multiplying the behavior motion complexity and the behavior confusion to obtain the behavior complexity degree of each dynamic region in the current frame of security monitoring image; obtaining the blur degree of each dynamic region in the current frame of security monitoring image according to the behavior complexity degree, comprising: obtaining the behavior complexity degree sequence of the i-th dynamic region in the current frame of security monitoring image; , a blur degree of the i-th dynamic region representing the current frame security monitoring image; a behavior complexity degree of the i-th dynamic region representing the current frame security monitoring image; a data number in a behavior complexity degree sequence of the i-th dynamic region representing the current frame security monitoring image; a (d+1)-th data value in the behavior complexity degree sequence of the i-th dynamic region representing the current frame security monitoring image; a d-th data value in the behavior complexity degree sequence of the i-th dynamic region representing the current frame security monitoring image; obtaining the enhancement coefficient of the current frame of security monitoring image according to the blur degree, enhancing the current frame of security monitoring image to obtain the current frame of security monitoring enhancement image.

2. The method of claim 1, wherein, The method comprises the following steps: presetting a difference threshold T, taking the absolute value of the gray difference between any pixel point in the current frame of security monitoring image and the pixel point at the same position in the previous frame of security monitoring image as a first difference, if the ratio of the first difference to 255 is greater than or equal to the difference threshold T, the pixel point in the current frame of security monitoring image is recorded as a dynamic pixel point; the region composed of all adjacent dynamic pixel points in the current frame of security monitoring image is recorded as a dynamic region, and each dynamic region of the current frame of security monitoring image is obtained.

3. The method of claim 1, wherein, The method comprises the following steps: obtaining the matching corner point of each corner point of each dynamic region in the current frame of security monitoring image; taking the Euclidean distance between the position coordinates of the j-th corner point of the i-th dynamic region in the current frame of security monitoring image and the position coordinates of the matching corner point as the moving distance of the j-th corner point; taking the angle between the direction of the line connecting the position coordinates of the j-th corner point of the i-th dynamic region in the current frame of security monitoring image and the position coordinates of the matching corner point and the horizontal direction as the moving direction of the j-th corner point; taking the moving distance, moving direction and coordinate position of the j-th corner point of the i-th dynamic region in the current frame of security monitoring image as the feature vector of the j-th corner point.

4. The method of claim 3, wherein, The method comprises the following steps: generating SIFT descriptor for the corner points of each dynamic region of each frame of security monitoring image, and then using feature matching algorithm to match the corner points with similar SIFT descriptor between the current frame of security monitoring image and the corresponding dynamic region in the previous frame of security monitoring image, to obtain the matching corner point of each corner point of each dynamic region in the current frame of security monitoring image.

5. The method of claim 1, wherein, The behavior complexity sequence of the i-th dynamic region of the current frame security monitoring image is obtained, including: A preset number of historical frame images G; a matching relationship of dynamic regions between the current frame security monitoring image and the previous G frame security monitoring images is obtained according to an optical flow method, so as to obtain a corresponding dynamic region of each dynamic region of the current frame security monitoring image in the previous G frame security monitoring images; and the behavior complexity of the i-th dynamic region of the current frame security monitoring image and the previous G frame security monitoring images is arranged according to an image acquisition sequence, so as to obtain the behavior complexity sequence of the i-th dynamic region of the current frame security monitoring image.

6. The method of claim 1, wherein, The current frame security monitoring image is enhanced by using the obtained enhancement coefficient of the current frame security monitoring image, so as to obtain a current frame security monitoring enhanced image. , an enhancement coefficient representing the current frame security monitoring image; a blur degree of the i-th dynamic region of the current frame security monitoring image; a mean value of the blur degrees of all dynamic regions of the current frame security monitoring image; the number of dynamic regions of the current frame security monitoring image; taking the enhancement coefficient of the current frame security monitoring image as the sharpening intensity of each pixel point in all dynamic regions of the current frame security monitoring image, and taking the sharpening intensity of each pixel point in the non-dynamic region of the current frame security monitoring image as 0; based on the sharpening intensity of each pixel point in the current frame security monitoring image, using a Laplacian sharpening algorithm to enhance the current frame security monitoring image, to obtain a current frame security monitoring enhanced image.

7. A machine vision based security monitoring image enhancement system, characterized by, The method comprises the steps of: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a machine vision-based security monitoring image enhancement method according to any one of claims 1-6 is realized.

Images