Security and protection monitoring big data processing system based on cloud computing

Abstract

The invention discloses a security and protection monitoring big data processing system based on cloud computing, and the system comprises a data receiving module which is used for receiving security and protection monitoring big data which comprises video data; relates to the technical field of data processing, through the arrangement of a target picture frame extraction module, video data are not directly transmitted, stored and processed after the video data are received, a target picture frame is extracted from mass video data, a specific area is determined by using a corner detection algorithm, and a target picture frame is obtained. A to-be-deleted picture group is divided through comparison values generated by comparing a specific area through a Bhattacharyya distance, a Hamming distance and the like, a large number of repeated or similar picture frames in the video data are deleted, then the target picture frame group is transmitted, stored and processed subsequently, the data volume of subsequent processing is greatly reduced for the video data, and the video data processing efficiency is improved. Calculation time and storage resources are effectively saved, and the overall processing efficiency of the video data is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a security monitoring big data processing system based on cloud computing. Background Art

[0002] Security monitoring big data refers to a large, diverse, complex and highly growing data set generated by various security monitoring devices in the field of security monitoring.

[0003] Currently, in the processing of security monitoring big data, video data processing faces many challenges. On the one hand, directly processing raw video data involves huge computing amounts and high storage costs. Videos contain a large amount of redundant information. Processing frame by frame is not only time-consuming and laborious, but also occupies a large amount of computing resources and storage space, resulting in low data processing efficiency and difficulty in meeting the requirements of real-time and high efficiency. On the other hand, when performing tasks such as image recognition, object detection and tracking, the complex background in video frames often seriously interferes with object analysis. For example, static background elements such as walls, desks and chairs in the monitoring scene will increase the difficulty of the algorithm to extract object features, reduce the accuracy of recognition and tracking, and it is impossible to accurately focus on key objects such as people, making it difficult to achieve precise analysis of people's actions and behaviors. Summary of the Invention

[0004] To solve the technical problems in the background art, the present invention proposes a security monitoring big data processing system based on cloud computing.

[0005] A security monitoring big data processing system based on cloud computing proposed by the present invention includes: A data receiving module: used to receive security monitoring big data, which includes video data; A target picture frame extraction module: select each video data from the video data. For multiple picture frames of the video data, take the first picture frame of the video data as the target picture frame; For the target picture frame, use a corner detection algorithm to find the corners in the target picture frame. After determining the corners, take the detected corners as the center and determine a fixed-size area as a specific area; Each corner corresponds to a specific area. After confirming the specific areas of the target picture frame, compare the multiple picture frames in the video data except the first picture frame with the target picture frame in sequence and enter the comparison mode. The comparison mode is used to divide the comparison picture frames into a picture group to be deleted, or retain the comparison picture frames and set the retained comparison picture frames as new target picture frames. If a new target picture frame is generated, the original target picture frame is classified into the target picture frame group; A background removal module: used to remove the background of each target picture frame in the target picture frame group; Image analysis module: Perform image recognition, object detection, and object tracking operations on multiple background-removed images through cloud computing.

[0006] Preferably, in the target picture frame extraction module, the comparison mode is as follows: When a picture frame is compared with the target picture frame, it is set as the comparison picture frame. For each specific area of the target picture frame, a comparison value is generated. And the comparison picture frame will mark a specific area at the same position and size as the target picture frame on itself. Then, a comparison value is generated for each specific area on the comparison picture frame. The comparison values corresponding to each specific area of the comparison picture frame and the target picture frame are compared. If the set conditions are met, the comparison picture frame is classified into the group of pictures to be deleted. Otherwise, the comparison picture frame is retained, and the retained comparison picture frame is used as the new target picture frame. Multiple subsequent picture frames in the video data are compared with the new target picture frame in sequence. The comparison of the target picture frame in this round ends, and the target picture frame in this round is classified into the target picture frame group; Preferably, in the comparison mode, when calculating the comparison value between the comparison picture frame and the target picture frame, it is as follows: S11: In the comparison picture frame and the target picture frame, for each specific area, scale the specific area to a fixed size. This step is to simplify subsequent calculations while retaining the main structural features of the image; S12: If the image of the specific area is a color image, calculate the color histogram of the scaled specific area color image, and use the Bhattacharyya distance to compare the color histograms of the corresponding specific areas in the picture frame and the target picture frame; Set the Bhattacharyya distance threshold. If the calculated Bhattacharyya distance is less than or equal to the set Bhattacharyya distance threshold, the picture frame being compared is classified into the group of pictures to be deleted. Otherwise, the picture frame being compared is retained, and the retained picture frame is used as the new target picture frame; The smaller the Bhattacharyya distance, the more similar the color distributions of the two images, and the higher the image similarity.

[0007] S12: If the image of the specific area is a black-and-white image, convert the scaled black-and-white image of the specific area into a grayscale image. Grayscale images are more conducive to subsequent feature extraction because they only focus on the brightness information of the image, and many structural features of the image are closely related to brightness changes; Perform a discrete cosine transform on the grayscale image. The DCT transform converts the image from the spatial domain to the frequency domain, highlighting the low-frequency components of the image. The low-frequency components contain the main structural information of the image, such as the general outline of the objects in the image, etc.; Take the low-frequency part of the DCT transformation result. The low-frequency coefficients represent the slowly changing parts of the image and are the main carriers of the image structure. The high-frequency coefficients correspond to the details and noise in the image and are discarded in this step to highlight the main structural features; Add up all the low-frequency coefficients and then divide by the total number of low-frequency coefficients to calculate the average value of the retained low-frequency coefficients; this average value is used as the reference value for subsequent comparisons; Compare each low-frequency coefficient with the average value. If it is greater than or equal to the average value, set it to 1; if it is less, set it to 0; arrange these 0s and 1s in order to form a hash value; Compare the hash values of the corresponding specific regions of the target picture frame and the comparison picture frame bit by bit; count the number of different bits in the two hash values and calculate the Hamming distance between the target picture frame and the comparison picture frame.

[0008] Set a Hamming distance threshold. If the Hamming distance is less than or equal to the set threshold, the comparison picture frame is classified into the picture group to be deleted; otherwise, the comparison picture frame is retained, and the retained picture frame is used as the new target picture frame; Preferably, in S12, the black and white image is converted into a grayscale image, and the conversion is carried out by the direct assignment method, the threshold method or the method of increasing the gray level.

[0009] Preferably, in the background removal module, the background of the target picture frame is removed as follows: For any target picture frame, select the picture frame with the highest similarity to it from the picture group to be deleted as the reference frame. The highest similarity means the smallest Bhattacharyya distance or the smallest Hamming distance. Each target picture frame corresponds to a reference frame, and the picture frames in the same picture group to be deleted can be selected as the reference frame by multiple target picture frames at the same time. Then, delete the picture frames in the picture group to be deleted except the reference frame; the most similar frame is used as the reference frame to improve the background subtraction effect Convert both the target picture frame and its corresponding reference frame into grayscale images. The purpose is to simplify the calculation process because the frame difference method can also effectively highlight the difference between the foreground and the background on grayscale images and the calculation amount is relatively small.

[0010] Perform a pixel-by-pixel subtraction operation on the gray values of the target picture frame and the reference frame to obtain a frame difference image: For each pixel point (x, y), calculate the difference between its gray value in the current target picture frame and the reference frame: ; where is the gray value of the pixel point (x, y) in the current target picture frame, is the gray value of the corresponding pixel point in the reference frame; The larger the obtained value of D(x, y), the greater the difference between the two frames at this pixel point, and the more likely it is to be the foreground.

[0011] Set a threshold T and compare the difference D(x, y) of each pixel point in the frame difference image with the threshold T: If D(x, y) > T, then determine this pixel point as a foreground pixel and set it to white (for example, grayscale value 255); If D(x, y) ≤ T, then determine it as a background pixel and set it to black (for example, grayscale value 0).

[0012] After the frame difference image undergoes threshold processing, a binary image is formed, and then morphological image processing is performed on the binary image; The combined use of dilation and erosion operations can effectively smooth the edges of foreground objects and improve the quality of foreground and background separation.

[0013] The binary image after morphological processing is used as a mask, which identifies the foreground area.

[0014] Perform a bitwise AND operation on the mask and the original current target picture frame. At this time, the part that is white (foreground) in the mask will retain the corresponding pixel values in the original target picture frame, while the part that is black (background) in the mask will become black, so as to obtain the target picture frame after removing the background. At this time, delete the reference frame corresponding to the target picture frame.

[0015] Realize the background removal of multiple target picture frames with different backgrounds; The bitwise AND operation to remove the background is a technical means used in image processing to separate the foreground and background.

[0016] Preferably, in the target picture frame extraction module, the corner detection algorithm adopts one of the Harris corner detection algorithm and the Shi - Tomasi corner detection algorithm.

[0017] The Bhattacharyya distance threshold for color images and the Hamming distance threshold for black and white images can be set and adjusted through experiments or experience according to the requirements of the security monitoring scenario and the image characteristics.

[0018] Preferably, in the background removal module, for the threshold T, an adaptive threshold algorithm is used to determine; The adaptive threshold algorithm adopts one of the mean adaptive threshold algorithm, the Gaussian adaptive threshold algorithm, and the Otsu algorithm.

[0019] A security monitoring big data processing method based on cloud computing includes the following steps: S1. Receive security monitoring big data and perform video frame extraction on the video data therein; For multiple picture frames of video data, first take the first picture frame of the video data as the target picture frame, use a corner detection algorithm to find the corners in the target picture frame, and determine a fixed-size area centered on the corners as a specific area; S2. Compare multiple picture frames in the video data except the first picture frame with the target picture frame in sequence and enter the comparison mode; Generate comparison values for each specific area of the target picture frame. The comparison picture frame delimits specific areas of the same position and size as the target picture frame and generates comparison values. Compare the comparison values of the corresponding specific areas of the two. If the set conditions are met, divide the comparison picture frame into the picture group to be deleted; otherwise, retain it and set it as the new target picture frame, and divide the original target picture frame into the target picture frame group; S3. If the specific area is a color image, calculate the color histogram after scaling and compare it using the Bhattacharyya distance. If it is less than or equal to the Bhattacharyya distance threshold, divide it into the selected group; otherwise, retain it. If it is a black-and-white image, convert it to a grayscale image and then perform discrete cosine transform, take the low-frequency part to calculate the average value, generate a hash value based on this, calculate the Hamming distance. If it is less than or equal to the Hamming distance threshold, divide it into the selected group; otherwise, retain it; S4. In the target picture frame group, each target picture frame selects the picture frame with the smallest Bhattacharyya distance or Hamming distance from the picture group to be deleted as the reference frame, and deletes the remaining picture frames in the picture group to be deleted; convert the target picture frame and the reference frame into grayscale images, subtract pixel by pixel to obtain a frame difference image, use an adaptive threshold algorithm to determine the threshold T, compare the difference D(x, y) of each pixel point with the threshold T to divide the foreground and background to form a binary image; perform morphological processing of dilation and erosion on the binary image, and use the processed binary image as a mask to perform bitwise AND with the original target picture frame to obtain the target picture frame after removing the background. At this time, delete the reference frame corresponding to the target picture frame; S5. Perform image recognition, target detection, and target tracking on multiple images after removing the background.

[0020] In the present invention, the proposed security monitoring big data processing system based on cloud computing has the following beneficial technical effects: 1. Through the setting of the target picture frame extraction module, after receiving the video data, it does not directly transmit, store, and process the video data. Instead, it extracts the target picture frame from the massive video data, uses the corner detection algorithm to determine the specific area, and then compares the comparison values generated by the specific area through the Bhattacharyya distance, Hamming distance, etc., divides the picture group to be deleted, deletes a large number of duplicate or similar picture frames in the video data, and then transmits, stores, and processes the target picture frame group later. For video data, it greatly reduces the amount of data to be processed later, effectively saves computing time and storage resources, and improves the overall processing efficiency of video data.

[0021] 2. The setting of the background removal module uses an adaptive threshold algorithm combined with morphological processing to separate the foreground and background of the target picture frame more accurately, which provides a high-quality pure image for subsequent image recognition, object detection, and tracking, helps improve the accuracy of related tasks, and strongly promotes the development and optimization of algorithms in the field of computer vision.

[0022] 3. In the comparison mode, when calculating the comparison value between the comparison picture frame and the target picture frame, for each specific region, the specific region is scaled to a fixed size to simplify subsequent calculations while retaining the main structural features of the image. Then, the Bhattacharyya distance or Hamming distance is calculated for the specific region. By only calculating for the specific region instead of the whole picture frame during the comparison value calculation, the amount of data processing is reduced, the computational complexity is lowered, and the specific region usually contains key information. Calculating only for these regions can focus more on the comparison of key information, avoid interference from irrelevant information such as the background, thereby improving the accuracy of the comparison and more accurately judging the similarity or difference between picture frames.

[0023] 4. In the security monitoring scenario, the setting of the background removal module can quickly separate people from the background, greatly facilitating the continuous tracking of specific individuals, promptly detecting abnormal behaviors and potential threats. At the same time, it effectively excludes background interference, making event detection more accurate, can quickly trigger the alarm mechanism, provides clear and accurate event information for security personnel, greatly improves the monitoring efficiency and security, and provides strong support for security work.

[0024] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. Brief Description of the Drawings

[0025] Figure 1 is the principle block diagram of the system of the present invention; Figure 2 is the flowchart of the method of the present invention. Detailed Description of the Embodiment

[0026] The embodiments of the present invention will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar symbols represent the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention.

[0027] As Figure 1 shown, a security monitoring big data processing system based on cloud computing includes: Data receiving module: used to receive security monitoring big data, which includes video data; Target Picture Frame Extraction Module: Select each video data from the video data. For multiple picture frames of the video data, take the first picture frame of the video data as the target picture frame; For the target picture frame, use a corner detection algorithm to find the corners in the target picture frame. After determining the corners, take the detected corners as the center and determine a fixed-size area as the specific area; Each corner corresponds to a specific area. After confirming the specific areas of the target picture frame, compare the multiple picture frames in the video data except the first picture frame with the target picture frame in sequence and enter the comparison mode. The comparison mode is used to divide the comparison picture frames into the picture group to be deleted, or retain the comparison picture frames and set the retained comparison picture frames as the new target picture frame. If a new target picture frame is generated, the original target picture frame is classified into the target picture frame group; In an optional embodiment, in the target picture frame extraction module, the corner detection algorithm adopts one of the Harris corner detection algorithm and the Shi - Tomasi corner detection algorithm.

[0028] Background Removal Module: Used to remove the background of each target picture frame in the target picture frame group; Image Analysis Module: Perform image recognition, object detection, and object tracking operations on the multiple images after removing the background through cloud computing.

[0029] In an optional embodiment, in the target picture frame extraction module, the comparison mode is as follows: When a picture frame is compared with the target picture frame, it is set as the comparison picture frame. For each specific area of the target picture frame, a comparison value is generated for each specific area. And the comparison picture frame will draw a specific area with the same position and size as the target picture frame on itself, and then a comparison value is generated for each specific area on the comparison picture frame. Compare the comparison values corresponding to each specific area of the comparison picture frame and the target picture frame. If the set conditions are met, the comparison picture frame is divided into the picture group to be deleted; otherwise, the comparison picture frame is retained, and the retained comparison picture frame is used as the new target picture frame. Compare the multiple picture frames following the new target picture frame in the video data with the new target picture frame in sequence. The comparison of the target picture frame in this round ends, and the target picture frame in this round is classified into the target picture frame group; Through the setting of the target picture frame extraction module, after receiving video data, instead of directly transmitting, storing, and processing the video data, it extracts the target picture frames from the massive video data, determines specific regions using corner detection algorithms, and then generates comparison values for the specific regions through comparison methods such as Bhattacharyya distance and Hamming distance. It divides the picture groups to be deleted, deletes a large number of duplicate or similar picture frames in the video data, and then transmits, stores, and processes the target picture frame groups subsequently. For video data, it greatly reduces the amount of data for subsequent processing, effectively saves computing time and storage resources, and improves the overall processing efficiency of video data.

[0030] In an optional embodiment, in the comparison mode, when calculating the comparison value between the comparison picture frame and the target picture frame, it is as follows: S11. In the comparison picture frame and the target picture frame, for each specific region, scale the specific region to a fixed size. This step is to simplify subsequent calculations while retaining the main structural features of the image. S12. If the image of the specific region is a color image, calculate the color histogram of the scaled color image of the specific region, and use the Bhattacharyya distance to compare the color histograms of the corresponding specific regions in the picture frame and the target picture frame. Set the Bhattacharyya distance threshold. If the calculated Bhattacharyya distance is less than or equal to the set Bhattacharyya distance threshold, the comparison picture frame is classified into the picture group to be deleted; otherwise, the comparison picture frame is retained, and the retained picture frame is used as the new target picture frame. The smaller the Bhattacharyya distance, the more similar the color distributions of the two images, and the higher the image similarity.

[0031] S12. If the image of the specific region is a black-and-white image, convert the scaled black-and-white image of the specific region into a grayscale image. Grayscale images are more suitable for subsequent feature extraction because they only focus on the brightness information of the image, and many structural features of the image are closely related to brightness changes. Perform a discrete cosine transform on the grayscale image. The DCT transform converts the image from the spatial domain to the frequency domain, highlighting the low-frequency components of the image. The low-frequency components contain the main structural information of the image, such as the approximate outline of objects in the image, etc. Take the low-frequency part of the DCT transform result. The low-frequency coefficients represent the slowly changing part of the image and are the main carriers of the image structure. The high-frequency coefficients correspond to the details and noise in the image and are discarded in this step to highlight the main structural features. Add up all the low-frequency coefficients and then divide by the total number of low-frequency coefficients to calculate the average value of the retained low-frequency coefficients; this average value is used as the reference value for subsequent comparisons. Compare each low-frequency coefficient with the average value. Those greater than or equal to the average value are set to 1, and those less than are set to 0. Arrange these 0s and 1s in order to form a hash value. For the hash values of the corresponding specific regions of the target picture frame and the comparison picture frame, compare them bit by bit. Count the number of different bits in the two hash values and calculate the Hamming distance between the target picture frame and the comparison picture frame.

[0032] Set a Hamming distance threshold. If the Hamming distance is less than or equal to the set threshold, the comparison picture frame is classified into the picture group to be deleted. Otherwise, keep the comparison picture frame, and use the kept picture frame as the new target picture frame. In an optional embodiment, in S12, the black and white image is converted into a grayscale image, and direct assignment method, threshold method or method of increasing gray levels is used for conversion.

[0033] In the comparison mode, when calculating the comparison value between the comparison picture frame and the target picture frame, for each specific region, scale the specific region to a fixed size to simplify subsequent calculations while retaining the main structural features of the image, and then calculate the Bhattacharyya distance or Hamming distance for the specific region. By only calculating for the specific region instead of the whole picture frame during the comparison value calculation, the data processing volume is reduced, the operation complexity is lowered, and the specific region usually contains key information. Only calculating for these regions can focus more on the comparison of key information, avoid interference from irrelevant information such as the background, thereby improving the accuracy of the comparison and more accurately judging the similarity or difference between picture frames.

[0034] In an optional embodiment, in the background removal module, the background of the target picture frame is removed as follows: For any target picture frame, select the picture frame with the highest similarity to it from the picture group to be deleted as the reference frame. The highest similarity means the smallest Bhattacharyya distance or the smallest Hamming distance. Each target picture frame corresponds to a reference frame, and the picture frames in the same picture group to be deleted can be selected as the reference frame by multiple target picture frames at the same time. Then delete the picture frames in the picture group to be deleted except the reference frame; the most similar frame is used as the reference frame to improve the background subtraction effect. Convert both the target picture frame and its corresponding reference frame into grayscale images, aiming to simplify the calculation process because the frame difference method can also effectively highlight the difference between the foreground and the background on grayscale images and the calculation amount is relatively small.

[0035] Perform a pixel-by-pixel subtraction operation on the grayscale values of the target picture frame and the reference frame to obtain a frame difference image: For each pixel point (x, y), calculate the difference between its grayscale values in the current target picture frame and the reference frame: ; where is the gray value of the pixel point (x, y) in the current target image frame, is the gray value of the corresponding pixel point in the reference frame; The larger the D(x,y) value obtained in this way, the greater the difference between the two frames at this pixel point, and the more likely it is to be the foreground.

[0036] Set a threshold T and compare the difference D(x,y) of each pixel point in the frame difference image with the threshold T: If D(x,y) > T, then determine this pixel point as a foreground pixel and set it to white (for example, gray value 255); If D(x,y) ≤ T, then determine it as a background pixel and set it to black (for example, gray value 0).

[0037] After the frame difference image is processed by the threshold, a binary image is formed, and then morphological image processing is performed on the binary image; The combined use of dilation and erosion operations can effectively smooth the edges of foreground objects and improve the quality of foreground and background separation.

[0038] The binary image after morphological processing is used as a mask, which identifies the foreground area.

[0039] Perform a bitwise AND operation on the mask and the original current target image frame. At this time, the white (foreground) part in the mask will retain the corresponding pixel values in the original target image frame, while the black (background) part in the mask will become black, so as to obtain the target image frame after removing the background. At this time, delete the reference frame corresponding to the target image frame.

[0040] Realize the background removal of multiple target image frames with different backgrounds; The bitwise AND operation to remove the background is a technical means used in image processing to separate the foreground and background.

[0041] The setting of the background removal module, using the adaptive threshold algorithm combined with morphological processing, makes the separation of the foreground and background of the target image frame more accurate. This provides high-quality pure images for subsequent image recognition, target detection and tracking, helps to improve the accuracy of related tasks, and strongly promotes the development and optimization of algorithms in the field of computer vision.

[0042] The Bhattacharyya distance threshold of the color image and the Hamming distance threshold of the black and white image can be set and adjusted according to the requirements of the security monitoring scenario and the image characteristics through experiments or experience.

[0043] In the background removal module, for the threshold T, an adaptive threshold algorithm is used to determine; The adaptive threshold algorithm adopts one of the mean adaptive threshold algorithm, the Gaussian adaptive threshold algorithm, and the Otsu algorithm.

[0044] In the security monitoring scenario, the setting of the background removal module can quickly separate people from the background, greatly facilitating the continuous tracking of specific individuals, promptly detecting abnormal behaviors and potential threats. At the same time, it effectively eliminates background interference, making event detection more accurate, can quickly trigger the alarm mechanism, provides clear and accurate event information for security personnel, greatly improves the monitoring efficiency and security, and provides strong support for security work.

[0045] Such as Figure 2 shown, a big data processing method for security monitoring based on cloud computing includes the following steps: S1. Receive big data of security monitoring, and perform video frame extraction on the video data therein; For multiple picture frames of the video data, first take the first picture frame of the video data as the target picture frame, use the corner detection algorithm to find the corners in the target picture frame, and determine a fixed-size area centered on the corners as the specific area; S2. Compare the multiple picture frames in the video data except the first picture frame with the target picture frame in sequence and enter the comparison mode; Generate comparison values for each specific area of the target picture frame, draw specific areas with the same position and size as the target picture frame in the comparison picture frame and generate comparison values, compare the comparison values of the corresponding specific areas of the two, and if the set conditions are met, divide the comparison picture frame into the picture group to be deleted, otherwise retain it and set it as the new target picture frame, and divide the original target picture frame into the target picture frame group; S3. If the specific area is a color image, calculate the color histogram after scaling, compare with the Bhattacharyya distance, and if it is less than or equal to the Bhattacharyya distance threshold, divide it into the selected group, otherwise retain it; if it is a black-and-white image, perform discrete cosine transform after converting it to a grayscale image, take the low-frequency part to calculate the average value, generate a hash value accordingly, calculate the Hamming distance, and if it is less than or equal to the Hamming distance threshold, divide it into the selected group, otherwise retain it; S4. In the target picture frame group, for each target picture frame, select the picture frame with the smallest Bhattacharyya distance or Hamming distance from the picture group to be deleted as the reference frame, and delete the remaining picture frames in the picture group to be deleted; convert the target picture frame and the reference frame to grayscale images, subtract pixel by pixel to obtain the frame difference image, use the adaptive threshold algorithm to determine the threshold T, compare the difference D(x, y) of each pixel point with the threshold T to divide the foreground and background to form a binary image; perform morphological processing of dilation and erosion on the binary image, and use the processed binary image as a mask to perform bitwise AND operation with the original target picture frame to obtain the target picture frame after background removal. At this time, delete the reference frame corresponding to the target picture frame; S5. Perform image recognition, object detection, and object tracking on multiple background-removed images.

[0046] This can significantly reduce the number of image frames that need to be compared in detail and improve the screening efficiency. Additionally, when the feature differences of image frames in different scenarios are large, the set conditions can be dynamically adjusted according to the source scenarios of the image frames (such as indoor, outdoor, day, night, etc.) to adapt to the image similarity judgment in different scenarios.

[0047] Meanwhile, the content not described in detail in this specification belongs to the prior art well-known to those skilled in the art.

[0048] In the embodiments provided by the present invention, it should be understood that the disclosed system or method can be implemented in other ways. For example, the above-described invention embodiments are merely illustrative. For example, the division of modules is only a logical function division, and there can be other division methods in actual implementation.

[0049] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0050] In addition, in each embodiment of the present invention, the functional modules can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module. The above-integrated modules can be implemented in the form of hardware or in the form of a combination of hardware and software functional modules.

[0051] For those skilled in the operation and maintenance field, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and can be implemented in other specific forms without departing from the basic characteristics of the present invention.

[0052] As described above, only the preferred specific implementation manners of the present invention are provided, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. A cloud computing-based security monitoring big data processing system, characterized in that: include: Data receiving module: used to receive security monitoring big data, including video data; Target picture frame extraction module: select each video data from the video data, and for multiple picture frames of the video data, take the first picture frame of the video data as the target picture frame; For the target image frame, a corner point detection algorithm is used to find the corner points in the target image frame. After the corner points are determined, a fixed-size area is determined as the specific area with the detected corner points as the center. Each corner point corresponds to a specific area. After confirming the specific area of ​​the target picture frame, multiple picture frames except the first picture frame in the video data are compared with the target picture frame in sequence, and a comparison mode is entered. The comparison mode is used to divide the compared picture frames into a picture group to be deleted, or to retain the compared picture frames and set the retained compared picture frames as new target picture frames. If a new target picture frame is generated, the original target picture frame is included in the target picture frame group; Background removal module: used to remove the background of each target picture frame in the target picture frame group; Image analysis module: Performs image recognition, target detection and target tracking operations on multiple images after background removal through cloud computing.

2. The cloud computing-based security monitoring big data processing system according to claim 1 is characterized in that: In the target picture frame extraction module, the comparison mode is: When a picture frame is compared with a target picture frame, it is set as a comparison picture frame, and a specific area of ​​the target picture frame, each specific area generates a comparison value, and the comparison picture frame will mark out a specific area of ​​the same position and size as the target picture frame on itself, and then each specific area on the comparison picture frame generates a comparison value, and the comparison values ​​corresponding to each specific area of ​​the comparison picture frame and the target picture frame are compared. If the set conditions are met, the comparison picture frame is divided into the picture group to be deleted, otherwise the comparison picture frame is retained, and the retained comparison picture frame is used as the new target picture frame, and multiple picture frames subsequent to the new target picture frame in the video data are compared with the new target picture frame in sequence, and the comparison of this round of target picture frames is completed, and the target picture frame of this round is divided into the target picture frame group.

3. The cloud computing-based security monitoring big data processing system according to claim 2 is characterized in that: In the comparison mode, the comparison value calculation between the comparison picture frame and the target picture frame is as follows: S11, comparing the image frame with the target image frame, and for each specific area, scaling the specific area to a fixed size; S12, if the image of the specific area is a color image, calculating a color histogram of the color image of the specific area after scaling, and using Bhattacharyya distance to compare the color histograms of the corresponding specific area in the picture frame and the target picture frame; A Bhattacharyya distance threshold is set. If the calculated Bhattacharyya distance is less than or equal to the set Bhattacharyya distance threshold, the image frame to be compared is divided into the image group to be deleted. Otherwise, the image frame to be compared is retained and the retained image frame is used as the new target image frame. S12, if the image of the specific area is a black and white image, convert the scaled black and white image of the specific area into a grayscale image, perform discrete cosine transform on the grayscale image, take the low-frequency part of the DCT transform result, add all the low-frequency coefficients, and then divide by the total number of the low-frequency coefficients to calculate the average value of the retained low-frequency coefficients; Compare each low-frequency coefficient with the average value, set it to 1 if it is greater than or equal to the average value, and set it to 0 if it is less than the average value; arrange these 0s and 1s in order to form a hash value; Compare the hash values ​​of the target image frame and the comparison image frame corresponding to the specific area bit by bit; count the number of different bits in the two hash values, and calculate the Hamming distance between the target image frame and the comparison image frame; A Hamming distance threshold is set. If the Hamming distance is less than or equal to the set threshold, the compared picture frame is divided into the picture group to be deleted. Otherwise, the compared picture frame is retained and the retained picture frame is used as the new target picture frame.

4. The cloud computing-based security monitoring big data processing system according to claim 3 is characterized in that: In S12, the black-and-white image is converted into a grayscale image by using a direct value assignment method, a threshold value method or a grayscale level increase method.

5. The cloud computing-based security monitoring big data processing system according to claim 1 is characterized in that: In the background removal module, the background of the target image frame is removed as follows: For any target picture frame, the target picture frame selects a picture frame with the highest similarity with it from the picture group to be deleted as a reference frame, each target picture frame corresponds to a reference frame, and picture frames in the same picture group to be deleted can be selected as reference frames by multiple target picture frames at the same time, and then the picture frames in the picture group to be deleted except the reference frame are deleted; Convert the target image frame and its corresponding reference frame into grayscale images; Subtract the grayscale value of the target image frame from the reference frame pixel by pixel to obtain the frame difference image: For each pixel (x, y), calculate the difference between its grayscale value in the current target image frame and the reference frame: ; in is the gray value of the pixel (x, y) in the current target image frame, is the gray value of the corresponding pixel in the reference frame; Set a threshold T and compare the difference D(x,y) of each pixel in the frame difference image with the threshold T: If D(x,y) > T, the pixel is determined as a foreground pixel and set to white; If D(x,y)≤T, it is determined to be a background pixel and set to black; The frame difference image is processed by threshold value to form a binary image, and then the binary image is processed by morphological image processing; The binary image after morphological processing is used as a mask, and the mask is bitwise ANDed with the original current target image frame. At this time, the white part of the mask will retain the corresponding pixel value in the original target image frame, and the black part of the mask will become black, thereby obtaining the target image frame after removing the background. At this time, the reference frame corresponding to the target image frame is deleted.

6. The cloud computing-based security monitoring big data processing system according to claim 1 is characterized in that: In the target image frame extraction module, the corner detection algorithm adopts one of the Harris corner detection algorithm and the Shi-Tomasi corner detection algorithm.

7. The cloud computing-based security monitoring big data processing system according to claim 5 is characterized in that: In the background removal module, the threshold T is determined by an adaptive threshold algorithm; The adaptive threshold algorithm adopts one of the mean adaptive threshold algorithm, Gauss adaptive threshold algorithm and Otsu algorithm.

8. The cloud computing-based security monitoring big data processing method according to any one of claims 1 to 7 is characterized in that: The following steps are involved: S1. Receive security monitoring big data and extract frames of the video data therein; For multiple picture frames of video data, firstly, the first picture frame of the video data is taken as the target picture frame, and the corner point detection algorithm is used to find the corner point in the target picture frame, and a fixed-size area is determined with the corner point as the center as the specific area; S2, comparing multiple picture frames except the first picture frame in the video data with the target picture frame in sequence, and entering a comparison mode; A comparison value is generated for each specific area of ​​the target picture frame, a specific area of ​​the same position and size as the target picture frame is demarcated from the comparison picture frame and a comparison value is generated, and the comparison values ​​of the corresponding specific areas of the two are compared. If the set conditions are met, the comparison picture frame is divided into the picture group to be deleted, otherwise it is retained and set as the new target picture frame, and the original target picture frame is divided into the target picture frame group; S3. If the specific area is a color image, calculate the color histogram after scaling, and compare it with Bhattacharyya distance. If it is less than or equal to the Bhattacharyya distance threshold, it is included in the selected group, otherwise it is retained; if it is a black and white image, convert it to a grayscale image and perform discrete cosine transform, take the low-frequency part to calculate the average value, generate a hash value based on it, calculate the Hamming distance, and if it is less than or equal to the Hamming distance threshold, it is included in the selected group, otherwise it is retained; S4, in the target picture frame group, each target picture frame selects a picture frame with the smallest Bhattacharyya distance or Hamming distance from the picture group to be deleted as a reference frame, and deletes the remaining picture frames in the picture group to be deleted; The target image frame and the reference frame are converted into grayscale images, and the frame difference image is obtained by pixel-by-pixel subtraction. The threshold T is determined by an adaptive threshold algorithm, and the difference D(x, y) of each pixel is compared with the threshold T to divide the foreground and background into a binary image; the binary image is subjected to morphological processing of dilation and erosion, and the processed binary image is used as a mask and bitwise ANDed with the original target image frame to obtain the target image frame after removing the background. At this time, the reference frame corresponding to the target image frame is deleted; S5. Perform image recognition, target detection and target tracking on multiple images after background removal.

Images