Camera cloud platform storage system and method

Abstract

The invention relates to the technical field of image data processing, in particular to a camera cloud platform storage system and method, and the system comprises a dynamic and static region division module which is used for extracting a distribution region of motion pixels based on continuous pictures collected by a camera, comparing the change range of brightness and color, and carrying out the calculation of the distribution region of the motion pixels; and determining the range of the dynamic region, identifying and marking other static regions, and generating a dynamic and static region division result. According to the method, the dynamic characteristics and the static characteristics of the picture content are recognized and distinguished, differential processing modes are adopted for the static area and the dynamic area respectively, inter-frame similarity calculation is utilized in the static area, only key frames are reserved, repeated data storage is reduced, and consumption of storage resources is reduced. In the dynamic area, the motion features are extracted and stored in a classified mode, the storage strategy is adjusted according to the motion frequency and the change value, and the storage quality and resource occupation are effectively balanced.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a camera cloud platform storage system and method. Background Art

[0002] The field of image data processing technology mainly revolves around image acquisition, transmission, storage, analysis and processing. This field involves computer vision, image compression, video encoding and decoding, artificial intelligence algorithms and other technologies. By processing image data, it helps to achieve functions such as target detection, pattern recognition, feature extraction, data storage and analysis. Image data processing is widely used in security monitoring, medical imaging, autonomous driving, smart cities, smart manufacturing and other industries.

[0003] The camera cloud platform storage system can upload the data collected by distributed camera devices to the cloud through the network for centralized storage and management, and supports real-time viewing, historical playback, intelligent analysis, and multi-terminal access. Its main uses include security monitoring, intelligent transportation, industrial monitoring and other scenarios, providing users with convenient storage and efficient image data processing capabilities, while reducing hardware equipment costs and improving data security and availability.

[0004] The existing technology has difficulty in accurately identifying the dynamic and static characteristics of the image content in the camera cloud platform, resulting in the use of consistent transmission and storage methods for static and dynamic images, and high data redundancy in the static area. For example, for monitoring scenes that have no obvious changes for a long time, repeated storage of complete frames takes up a lot of storage space, increasing system costs. For dynamic areas, the existing technology fails to classify and process according to the frequency of motion or the range of change, resulting in confusion in the priority of high-frequency dynamic images in transmission and storage, and unreasonable resource allocation. When network conditions are poor, failure to adjust the data transmission strategy in real time or record the specific data of incomplete transmission may result in loss of dynamic scene data or serious transmission delays, further reducing data integrity and user experience. For example, in the face of sudden dynamic changes in the monitoring screen, network congestion may cause key frames to fail to be transmitted in time, affecting subsequent data analysis and judgment. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a camera cloud platform storage system and method.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A camera cloud platform storage system includes:

[0007] The dynamic and static area division module compares the range of brightness and color changes based on the continuous images captured by the camera, extracts the distribution area of ​​motion pixels, determines the range of the dynamic area for the distribution area of ​​motion pixels, and identifies and marks the remaining static areas to generate dynamic and static area division results;

[0008] The picture repetition evaluation module extracts pixel values ​​in the static area based on the dynamic and static area division results, obtains histogram data of the static area, calculates the similarity between the histogram data of the static area, marks the static area and repeated pictures according to the similarity, records the frequency and change value of the dynamic motion area, and generates image data of picture repetition and dynamic features;

[0009] The partition storage optimization module classifies the static area according to the inter-frame repetition based on the image data of the picture repetition and dynamic characteristics, and classifies the dynamic area according to the frequency and change value of the motion vector, performs partition compression storage, and obtains optimized regional storage records;

[0010] The dynamic transmission and feedback module sorts the compressed data of the dynamic and static areas according to the priority of the regional storage data based on the optimized regional storage records, optimizes the transmission of image data, and generates camera storage image transmission data record results.

[0011] As a further solution of the present invention, the step of acquiring the distribution area of ​​the motion pixels is specifically:

[0012] Based on the continuous picture data collected by the camera, the brightness and color values ​​of the image are extracted according to each pixel of the image frame, and the brightness and color changes of each pixel in the continuous frames are compared and analyzed to obtain the change amplitude and direction information of the pixel points;

[0013] Based on the change amplitude and direction information of the pixel points, a set of pixel points whose changes exceed a threshold is determined, the area where the pixel points are located is extracted and marked, and the boundary of the dynamic area is divided to obtain the distribution area information of the moving pixels.

[0014] As a further solution of the present invention, the steps of obtaining the dynamic and static area division results are specifically as follows:

[0015] Based on the distribution area information of the motion pixels, densely distributed pixels are aggregated to form the dynamic area boundary by analyzing the adjacency relationship of the pixels, sparsely distributed pixels are removed, and static areas are marked by combining the brightness mean and the frequency of inter-frame changes to generate preliminary division data of dynamic and static areas;

[0016] Based on the preliminary division data of the dynamic and static areas, the formula is adopted:

[0017]

[0018] Calculate the difference P after smoothing at the boundary between the dynamic and static areas, process the boundary between the dynamic and static areas according to the difference P, and obtain the dynamic and static area division results;

[0019] Among them, n is the total number of pixels at the junction of the dynamic and static areas, B i represents the gray value of the i-th dynamic boundary pixel, S i Represents the grayscale value of the i-th static boundary pixel.

[0020] As a further solution of the present invention, the steps for obtaining the similarity between the histograms are specifically as follows:

[0021] Based on the dynamic and static area division results, the brightness values ​​and color values ​​of all pixels in the static area are extracted, the pixels are grouped and counted according to the brightness and color channels, and the number of pixels in each group is recorded to generate histogram data of the static area;

[0022] For the histogram data of the static area, the formula is used:

[0023]

[0024] Calculate the histogram similarity value G to obtain the similarity information between the histograms;

[0025] Where j represents the jth interval in the histogram, n′ is the total number of intervals in the histogram, and H Aj is the height value of the histogram of frame A in the jth interval, H Bj is the height value of the histogram of frame B in the jth interval, min(H Aj ,H Bj ) is the minimum value of the histogram height values ​​of frame A and frame B in the jth interval.

[0026] As a further solution of the present invention, the steps for acquiring the image data of the picture repetitiveness and dynamic characteristics are specifically as follows:

[0027] According to the similarity information between the histograms, by analyzing the similarity values ​​of the static areas frame by frame, marking the static areas above the image division threshold as high-similarity areas, marking the pictures to which the high-similarity areas belong as high-repetition pictures, and generating marking information of the high-repetition pictures;

[0028] Based on the motion vector data of the moving pixels in the dynamic area, the frequency and change of the moving area are counted by analyzing the motion direction, distance and brightness change between frames, and the motion characteristics within each dynamic area are extracted, including the average frequency and change range. Combined with the marking information of the high-repetition picture, image data of the picture repetition and dynamic characteristics are obtained.

[0029] As a further solution of the present invention, the step of obtaining the optimized regional storage record is specifically as follows:

[0030] Based on the image data of the picture repetition and dynamic characteristics, static areas are classified according to the inter-frame repetition, high-repetition areas and low-repetition areas are determined according to the repetition, high-repetition areas are extracted and their key frames are directly stored, and the boundary positions of the areas are recorded at the same time. For low-repetition areas, pixel values ​​are extracted frame by frame and encoded and compressed and stored in frame sequence to generate a partition storage result of repetition;

[0031] Based on the motion vector data of the dynamic area, classification is performed according to the motion frequency and change value. The area with a marking frequency lower than the set frequency threshold is marked as a low-change dynamic area and its storage bit rate is reduced. For the dynamic area with a frequency higher than the frequency threshold, its motion change characteristics are extracted and combined with the position record, combined with the partition storage result of the repetitiveness, to generate an optimized regional storage record.

[0032] As a further solution of the present invention, the steps of obtaining the result of recording the image transmission data stored by the camera are specifically as follows:

[0033] Based on the optimized regional storage records, the formula is adopted:

[0034]

[0035] Calculate the dimensionless priority F representing the i′th region i′ , get regional priority information;

[0036] Among them, C i′ is the data variation coefficient of the i′th region, C max is the maximum value of the coefficient of variation of data in all regions, R i′ is the data update frequency of the ith region, R max is the maximum value of the update frequency of all regional data, L i′ is the storage load factor of the ith region, L max is the maximum value of the storage load factor;

[0037] According to the regional priority information, the transmission capacity is allocated in real time according to the network bandwidth situation, and the camera storage image transmission data record result is obtained.

[0038] A camera cloud platform storage method, which is executed based on the camera cloud platform storage system, comprises the following steps:

[0039] S1: Based on the camera capture image, pixel brightness and color values ​​are extracted frame by frame, moving pixels are marked by inter-frame comparison, moving pixel areas are aggregated to define the boundary range, and the brightness change amplitude and frequency characteristics of non-moving areas are analyzed to mark static areas, generating dynamic and static area division results;

[0040] S2: Based on the dynamic and static area division results, extract the pixel brightness value and color value of the static area to generate histogram data, mark the high similarity area by comparing the histogram similarity values ​​frame by frame, extract the motion vector data from the dynamic area to count the frequency and change amplitude, and generate the picture repetition and dynamic feature image data;

[0041] S3: Based on the picture repetition and dynamic feature image data, extract key frames or compress and store them according to the repetition classification of static areas, identify and classify dynamic areas according to the motion vector frequency and change amplitude, and generate optimized area storage records;

[0042] S4: Based on the optimized regional storage records, dynamically adjust the regional priority order, transmit key data and feature points to high-priority regions, preferentially transmit dynamic region data, selectively transmit feature values ​​to low-priority static regions, and generate priority-ordered image transmission data;

[0043] S5: Based on the priority-sorted image transmission data, the transmission capacity is allocated in combination with the network bandwidth, high-priority data is transmitted first, the number and range of the untransmitted data frame in the low-priority area is paused or recorded, and the camera stores the image transmission data recording results.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are:

[0045] In the present invention, by identifying and distinguishing the dynamic and static characteristics of the picture content, differentiated processing methods are adopted for the static area and the dynamic area respectively. In the static area, the inter-frame similarity calculation is used to retain only the key frames, thereby reducing the storage of duplicate data and reducing the consumption of storage resources. For example, for scenes without obvious changes, redundant frames can be effectively eliminated to reduce storage pressure. The dynamic area extracts motion features and stores them in categories, and adjusts the storage strategy according to the motion frequency and change value. The high-frequency dynamic area is stored in the form of complete data, while the low-frequency dynamic area is stored at a reduced bit rate, which effectively balances the storage quality and resource occupancy. In addition, during the transmission process, data is scheduled according to the priority of the picture content, and the high-priority dynamic area is transmitted first to ensure real-time requirements. At the same time, only simplified feature values ​​are transmitted for the low-priority static area to reduce the transmission traffic. In the case of network abnormalities, the number and range of the dynamic data and static data that have not been transmitted are recorded to ensure the recoverability of subsequent transmission and avoid data loss. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a system flow chart of the present invention;

[0047] Figure 2 A flow chart of obtaining a distribution area of ​​motion pixels according to the present invention;

[0048] Figure 3 A flow chart of obtaining dynamic and static area division results of the present invention;

[0049] Figure 4 A flow chart of calculating the similarity between histograms of the present invention;

[0050] Figure 5 A flow chart of the present invention for obtaining image data of picture repetitiveness and dynamic characteristics;

[0051] Figure 6 A flowchart for obtaining optimized regional storage records for the present invention;

[0052] Figure 7 The present invention is a flow chart of obtaining the camera storage image transmission data recording result. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0054] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0055] See also Figure 1 , the present invention provides a technical solution: a camera cloud platform storage system includes.

[0056] The dynamic and static area division module obtains the brightness and color values ​​of each pixel based on the continuous images captured by the camera, compares the brightness and color change range frame by frame, marks the pixels whose changes exceed the preset image division threshold, extracts the distribution area of ​​motion pixels, and determines the range of the dynamic area through boundary aggregation operation for the distribution area of ​​motion pixels. For the part not marked as dynamic area, the static area is identified and marked by analyzing the brightness mean value in the frame and the frequency of change between frames. The boundary between dynamic and static areas is processed by boundary smoothing calculation to generate dynamic and static area division results.

[0057] The picture repetition evaluation module extracts the pixel values ​​in the static area based on the dynamic and static area division results, constructs the brightness and color histograms of each frame according to the color channel, obtains the histogram data of the static area, calculates the similarity between the histogram data of the static area, judges the inter-frame repetition by comparing the similarity value with the image division threshold, and marks the static area with a similarity higher than the preset similarity reference value as a high-similarity static area and a high-repetition picture according to the preset similarity reference value. For the dynamic area, the frequency and change value of the motion area are recorded according to the motion vector and change range of the motion pixel, and the image data of the picture repetition and dynamic characteristics are generated;

[0058] The partition storage optimization module classifies static areas according to the inter-frame repetition and dynamic characteristics of the image data, and classifies dynamic areas according to the frequency and change value of the motion vector, and implements partition compression storage, including directly storing high-repetition frames as key frames for static areas in high-repetition images according to the inter-frame repetition threshold of the static area, and recording the position of the high-repetition area. For static areas in low-repetition images, the pixel values ​​of each frame are obtained, and encoded and compressed and stored in frame sequence. For dynamic areas, according to the motion frequency and change value recorded by the motion vector, the dynamic areas with a frequency lower than the set frequency threshold are identified, and these areas are marked as low-change dynamic areas to reduce their storage bit rate. For dynamic areas with higher frequencies, the optimized regional storage records are obtained by extracting the change value and range of the motion vector, combining the marked dynamic area position, and fully recording its motion characteristics;

[0059] The dynamic transmission and feedback module is based on the optimized regional storage records and sorts the compressed data of the dynamic and static areas according to the priority of the regional storage data. Combined with the current camera cloud platform storage network bandwidth, it obtains the current transmission capacity in real time, and selects the motion change data and key feature points of the high-priority dynamic area for priority transmission through priority scheduling. The transmission frequency of the low-priority static area data is reduced, or only the histogram feature value is transmitted. Under abnormal network conditions, the frame number and data range of the dynamic area or static area that are not transmitted are recorded to generate the camera storage image transmission data record results.

[0060] The dynamic and static area division results include the motion pixel set, motion boundary range and boundary smoothing processing data of the dynamic area, the brightness mean distribution, inter-frame change frequency and area marking information of the static area, the image data of the picture repetition and dynamic characteristics include the inter-frame repetition classification mark of the static area, the similarity data and feature classification mark of the high-repetition static area, and the motion frequency data, change value distribution and frequency classification label of the dynamic area. The optimized area storage records include the key frame index and storage location of the high-repetition static area, the compressed storage data of the low-repetition static area, the reduced bit rate storage records of the low-frequency dynamic area, and the motion change compression data and partition feature information of the high-frequency dynamic area. The camera storage image transmission data record results include the motion change transmission record of the high-priority dynamic area, the key feature point transmission data, the compressed data transmission record of the low-priority static area, the unsent frame number of the abnormal transmission, the frame loss time range and data mark.

[0061] See also Figure 2 , the steps for obtaining the distribution area of ​​motion pixels are specifically as follows:

[0062] Based on the continuous picture data collected by the camera, the brightness and color values ​​of the image are extracted according to each pixel of the image frame, and the brightness and color changes of each pixel in the continuous frames are compared and analyzed to obtain the change amplitude and direction information of the pixel points;

[0063] Based on the continuous images captured by the camera, the data is uploaded to the cloud platform through the network for centralized storage. The image frames stored in the cloud platform are processed using OpenCV. The brightness and color values ​​of the pixels in each frame are read from the stored image sequence. The image data is loaded and converted into a pixel matrix. The brightness value is extracted by converting the RGB image into a grayscale image, and the color value is directly obtained from the matrix data of the RGB channel. The cloud platform obtains the change amplitude and direction of each pixel by comparing the brightness and color changes of each pixel in the continuous frames. For example, for each pixel in two consecutive frames, the absolute difference of its brightness value is calculated, and it is determined whether it exceeds the set threshold. For example, for a certain pixel in frame A and frame B, if the brightness of frame A is 120 and the brightness of frame B is 110, the change amplitude is |120-110|=10. If this amplitude value is greater than the threshold (such as 8), the pixel is recorded as a change point. By calculating the difference in the color values ​​of the pixels in two consecutive frames, the direction is determined by the maximum component of the difference in the RGB channels after the change. For example, for the color values ​​of frame A (100, 120, 90) and frame B (105, 125, 100), the differences of the R, G, and B channels are calculated as |105-100|=5, |125-120|=5, |100-90|=10, respectively, and the direction of change is recorded as "increase B component". The magnitude and direction of the change are stored in a new matrix for analysis. According to the results of each inter-frame comparison, the magnitude and direction of the change are uploaded to the cloud database for unified management. After the processing is completed, the system will provide basic pixel data for subsequent marking of the changed area.

[0064] Based on the change amplitude and direction information of the pixel points, determine the set of pixel points whose changes exceed the threshold, extract and mark the areas where these pixel points are located, and divide the boundaries of the dynamic area to obtain the distribution area information of the moving pixels;

[0065] Pixels whose changes exceed the preset threshold are marked, and the contour recognition method in OpenCV is used to complete the extraction of marked areas in the cloud platform processing environment. First, the set of pixels exceeding the threshold is determined by comparing matrix data. These sets form binary marked images, which are uploaded to the cloud for further analysis of dynamic areas. The distribution range of pixels in the marked image is processed by extracting contours, and the distribution of motion areas is determined by extracting boundaries. K-means clustering is used to further classify these pixels, and different dynamic area boundaries are identified according to the distribution characteristics of pixel points. The generated boundary information is stored as area coordinate data, and the final distribution results of the dynamic area are stored in cloud files in image format, thereby achieving accurate recognition and division of dynamic areas based on changing pixels.

[0066] See also Figure 3, the specific steps for obtaining the dynamic and static area division results are:

[0067] Based on the distribution area information of motion pixels, densely distributed pixels are aggregated to form the dynamic area boundary by analyzing the adjacency relationship of pixels, sparsely distributed pixels are removed, and static areas are marked by combining the brightness mean and inter-frame change frequency to generate preliminary division data of dynamic and static areas;

[0068] First, the spatial distribution of the extracted motion pixel data is analyzed. The boundary aggregation method is used to remove sparsely distributed pixels from the concentrated area. By detecting the brightness change amplitude and adjacency relationship of the pixels around each pixel, the pixels that meet the dense distribution standard are integrated into a complete dynamic area. At the same time, the boundary position is recorded, and the pixels that are not divided into the dynamic area are processed independently. These pixels are analyzed by the average brightness value within the frame, combined with the statistics of the change frequency of continuous frames, and the pixels with stable brightness changes and a frequency lower than the specified threshold are screened out and marked as static pixel areas. For example, the camera continuously stores the picture data on the cloud platform at 25 frames per second, and the distribution area of ​​the motion pixels is extracted through the data processing process. Assume that the motion pixel distribution data at a certain moment has been stored as a two-dimensional array, where the value "1" represents a motion pixel and the value "0" represents a non-motion pixel. When performing boundary aggregation processing on these motion pixels, the number of "1" around each motion pixel is first calculated according to the adjacency relationship to determine whether the pixel belongs to the dense distribution area. If more than 50% of the pixels in a motion pixel and its surrounding eight neighborhoods are motion pixels, it will be classified as a dynamic area. Use aggregation algorithms (such as connectivity analysis) to merge adjacent dynamic areas to form complete area boundaries. The boundary data is further stored as polygon coordinates for subsequent image analysis and segmentation tasks. For the parts that are not marked as dynamic areas (pixels with a value of "0"), the intra-frame brightness mean and inter-frame brightness change frequency are analyzed point by point. The following are the specific operations: Brightness mean: Extract the grayscale value of each pixel from the original frame data, calculate the average brightness of the small area (such as a 3×3 window) where it is located, and determine whether the pixel is in a stable brightness range (such as a fluctuation of less than ±5%). Inter-frame change frequency: By comparing the brightness value change records of the pixel in consecutive frames, count the number of changes in the brightness value in a specified time window (such as within 1 second). If the change frequency is lower than the threshold (such as 2 times / second), it is determined to be a static area. Pixels that meet the above conditions are marked as static areas, and the area segmentation result file in the cloud platform storage is updated.

[0069] Based on the preliminary division data of dynamic and static areas, the formula is used:

[0070]

[0071] Calculate the difference P after smoothing at the boundary between the dynamic and static areas, process the boundary between the dynamic and static areas according to the difference P, and obtain the dynamic and static area division results;

[0072] Among them, P is used to represent the brightness difference between dynamic and static pixels in the boundary area, n is the total number of pixels at the boundary between dynamic and static areas, and B i represents the gray value of the i-th dynamic boundary pixel. The brightness value of the boundary pixel is extracted after the image grayscale conversion. For example, the brightness data is directly obtained using the pixel intensity value in the grayscale image. i Represents the grayscale value of the i-th static boundary pixel. The brightness value of the corresponding boundary pixel is extracted through the image data marked with the static area.

[0073] For example, if there are 4 pixels at the junction of the dynamic and static areas of an image frame, the dynamic area boundary pixel brightness value set is 120, 125, 118, 130, and the static area boundary pixel brightness value set is 122, 128, 117, 129. The calculation process is as follows:

[0074]

[0075]

[0076]

[0077] The result shows that the brightness difference at the junction of the dynamic and static areas is 3.75 (relative brightness unit). By comparing the historical value or the set benchmark value (such as the benchmark value is 5), it can be concluded that this difference is low, indicating that the brightness change at the junction is smoother. Further, the result can be optimized through boundary smoothing to ensure that the division of dynamic and static areas is more continuous and natural, thereby improving the accuracy and visual effect of area division.

[0078] See also Figure 4 , the specific steps for calculating the similarity between histograms are:

[0079] Based on the results of dynamic and static area division, the brightness and color values ​​of all pixels in the static area are extracted, the pixels are grouped and counted according to the brightness and color channels, and the number of pixels in each group is recorded to generate the histogram data of the static area;

[0080] First, load the stored frame image data from the cloud platform, extract the pixels marked as static areas, and filter out the pixel values ​​of each channel in the static area according to the RGB color channels of the frame image. By counting the brightness and color values ​​of each pixel, these values ​​are divided into different intensity intervals, and the number of pixels in each interval is recorded. The brightness histogram is obtained by converting the grayscale values ​​of the pixels in the static area from the original frame data. Each grayscale value is grouped and counted to form columnar data. The color histogram corresponds to the three RGB channels respectively, and is completed by analyzing the intensity distribution of pixel values ​​in the channel separately. Finally, the brightness and color histograms of the static area of ​​each frame are stored in the cloud platform database in the form of arrays to form static area histogram data.

[0081] For the histogram data in the static area, the formula is used:

[0082]

[0083] Calculate the histogram similarity value G to obtain the similarity information between the histograms;

[0084] Among them, G is used to represent the similarity of the brightness or a color channel histogram between two frames, and the value range is 0 to 1. The closer the value is to 1, the higher the similarity. j represents the jth interval in the histogram, which is generated by grouping the brightness or color values. For example, the brightness values ​​are divided into 0-15, 16-31, etc., or the values ​​of each color channel are divided into intervals to generate indexes. n′ is the total number of histogram intervals, and H is the total number of histogram intervals. Aj is the height value of the histogram of frame A in the jth interval, obtained by counting the brightness value or color channel value of the static area in frame A, H Bj is the height value of the histogram of frame B in the jth interval, which is obtained by counting the brightness value or color channel value of the static area in frame B. min(H Aj ,H Bj ) is the minimum value of the histogram height values ​​of frame A and frame B in the jth interval, indicating the overlapping part of the pixel values ​​of the two frames in this interval.

[0085] For brightness histogram: The static area brightness histogram data of frame A and frame B is divided into 4 intervals: 10, 20, 30, 40 and 15, 25, 20, 35.

[0086] Calculate the overlap:

[0087] min(H A1 ,H B1 )=min(10,15)=10

[0088] min(H A2 ,H B2 )=min(20,25)=20

[0089] min(H A3 , H B3 )=min(30,20)=20

[0090] min(H A4 ,H B4 )=min(40,35)=35

[0091] Accumulate and calculate similarity:

[0092]

[0093] For color histogram (taking R channel as an example):

[0094] The red channel histogram data of the static area of ​​frame A and frame B are divided into 4 intervals, namely 8, 18, 28, 38 and 10, 20, 25, 30.

[0095] Calculate the overlap:

[0096] min(H A1 , H B1 )=min(8,10)=8

[0097] min(H A2 ,H B2 )=min(18,20)=18

[0098] min(H A3 ,H B3 )=min(28,25)=25

[0099] min(H A4 ,H B4 )=min(38,30)=30

[0100] Accumulate and calculate similarity:

[0101]

[0102] The results show that the similarities between the brightness histogram and the color histogram are 0.85 and 0.88 respectively.

[0103] See also Figure 5 , the steps for acquiring the image data of the picture repetitiveness and dynamic characteristics are as follows:

[0104] According to the similarity information between the histograms, by analyzing the similarity values ​​of the static areas frame by frame, the static areas above the image division threshold are marked as high-similarity areas, and the pictures belonging to the high-similarity areas are marked as high-repetition pictures, thereby generating marking information of the high-repetition pictures;

[0105] The inter-frame repetition is determined by comparing the similarity value with the image division threshold. The specific operation is as follows: first, the static area histogram data of the frame is loaded from the cloud platform, the similarity of the brightness histogram and the color histogram is calculated respectively, and the similarity value between each frame pair is recorded. Then, the calculated similarity value is compared with the preset image division threshold (for example, 0.9) to determine whether the frame is a high-repetition picture. For example, for the brightness histogram of frame A and frame B, the calculated similarity value is 0.85, which is lower than the image division threshold of 0.9, so frame B is not marked as a high-repetition picture; for the red channel color histogram of frame C and frame D, the similarity value is 0.88, which is also lower than the threshold, and frame D is not marked as a high-repetition picture. However, when the similarity of the blue channel color histogram of frame E and frame F is 0.92, which is higher than the threshold, frame F is marked as a high-repetition picture. All static area marking information with high similarity will be integrated into the marking file of cloud storage as the basis for judging the inter-frame repetition. Finally, the marked high-repetition frames and high-similarity static regions are used for further analysis and processing.

[0106] Based on the motion vector data of the moving pixels in the dynamic area, the frequency and change of the moving area are counted by analyzing the motion direction, distance and brightness change between frames, and the motion characteristics in each dynamic area are extracted, including the average frequency and change range. Combined with the marking information of the high-repetitive images, the image data of the image repetitiveness and dynamic characteristics are obtained;

[0107] For dynamic areas, the frequency and change value of the moving area are recorded by extracting the motion vector and change range of the moving pixels. The specific operation is as follows: First, the position coordinates of the marked moving pixels in each frame of the dynamic area are extracted, and the motion vector, including the direction and distance of movement, is calculated by comparing consecutive frames. For example, if a pixel in a frame moves from position (10,15) to position (12,18), the direction of its motion vector is 45 degrees and the distance is 3 pixels. Then, the motion vectors of all pixels in the entire dynamic area are counted to obtain the average motion frequency (such as 30 times per second) and motion change value (such as the brightness fluctuation range is ±20 gray levels) of the area. For example, in a group of dynamic areas, the average motion vector of the pixels in a certain area is 5 pixels to the right per frame, the motion frequency is 25 times per second; the brightness change range is from 100 to 120 gray levels. After recording the above statistical values, these data are integrated into an image data format to represent the picture repetitiveness and dynamic characteristics for subsequent storage and analysis. For example, areas with high frequency and large change values ​​may represent fast-moving objects, while areas with low frequency and low change may represent slower background movement. Ultimately, the generated image data will be stored on the cloud platform, providing complete information on image repeatability and dynamic features.

[0108] See also Figure 6, the steps for obtaining the optimized regional storage records are as follows:

[0109] Based on the image data of picture repetitiveness and dynamic characteristics, static areas are classified according to the repetitiveness between frames, and high repetitiveness areas and low repetitiveness areas are judged according to the repetitiveness. High repetitiveness areas are extracted and their key frames are directly stored, and the boundary positions of the areas are recorded at the same time. For low repetitiveness areas, pixel values ​​are extracted frame by frame and encoded and compressed and stored in frame sequence to generate partition storage results of repetitiveness;

[0110] Based on the image data of picture repetition and dynamic features stored in the camera cloud platform, the static area is classified according to the inter-frame repetition. First, the static area histogram similarity and inter-frame repetition information of each frame are extracted. For example, in a set of continuous frames, the static area similarity of frame A and frame B is 0.92, and the similarity of frame B and frame C is 0.89. For static areas whose inter-frame similarity is higher than the set repetition threshold (such as 0.9), they are marked as high repetition areas, and frame B is stored as a key frame. At the same time, the boundary position and coordinate data of the static area are recorded for subsequent processing. For frame C, since the similarity is lower than the repetition threshold, it is classified as a low repetition picture, and the pixel values ​​of its static area are extracted frame by frame. The encoding method (such as differential encoding) is used to compress and store them in sequence according to the frame order, and the encoding results are stored in the cloud platform database. For example, assuming that the repetition of the static area of ​​frames 1 to 4 in 10 consecutive frames is 0.93, 0.91, 0.88, and 0.85 respectively, the static area of ​​frames 1 to 2 is marked as high repetition, and frame 2 is stored as a key frame. At the same time, the coordinate range of the static area is recorded as (50, 50)-(200, 200). For the static area of ​​frames 3 to 4, because its repetition is lower than the repetition threshold, the pixel values ​​are extracted frame by frame, and the frame sequence data is encoded and stored using a compression algorithm. Finally, after the classification is completed, the storage information of the key frame and the compressed frame is synchronized to the cloud platform storage system to form an optimized static area partition storage result.

[0111] Based on the motion vector data of the dynamic area, the motion frequency and change value are classified. The area with a frequency lower than the set frequency threshold is marked as a low-change dynamic area and its storage bit rate is reduced. For the dynamic area with a frequency higher than the frequency threshold, its motion change characteristics are extracted and combined with the position record and the partition storage result of the repetition degree to generate an optimized regional storage record;

[0112] Based on the dynamic feature data of the picture stored in the camera cloud platform, the dynamic area is classified according to the motion frequency and change value recorded by the motion vector. First, the motion frequency of the moving pixels in each dynamic area is extracted. For example, by comparing the motion vectors of the pixels in consecutive frames, the motion frequency is calculated to be 20 times per second; at the same time, the motion change value is recorded, such as the brightness change range of ±15 gray levels. Then, the dynamic area is classified according to the set frequency threshold (such as 15 times / second): if the motion frequency is lower than the frequency threshold, the area is marked as a low-change dynamic area, and the storage space requirement is reduced by adjusting the storage bit rate; if the motion frequency is higher than the frequency threshold, the change value and range of the motion vector are further extracted, and the motion characteristics are fully recorded in combination with the marked position of the dynamic area. For example, suppose the dynamic area of ​​a frame is divided into two parts. The motion frequency of area A is 10 times / second, and the brightness variation range is ±10, which is lower than the set frequency threshold. Therefore, area A is marked as a low-variation dynamic area, and the storage bit rate is adjusted to 70% of the standard value; the motion frequency of area B is 25 times / second, and the brightness variation range is ±25, which is higher than the frequency threshold. The direction and speed change value of its motion vector are extracted, and the marked position of area B in the frame is recorded as (30, 30)-(100, 100). Finally, the classified dynamic area storage records will be uploaded to the cloud platform database to form an optimized area storage result.

[0113] See also Figure 7 , the specific steps for obtaining the camera storage image transmission data record result are:

[0114] Based on the optimized regional storage records, the formula is used:

[0115]

[0116] Calculate the dimensionless priority F representing the i′th region i′ , get regional priority information;

[0117] Among them, F i′ It is used to sort the dynamic and static areas. The higher the priority of the area, the earlier the transmission order. i′ is the data change coefficient of the i′th region, which is used to measure the importance of the data. For dynamic regions, it is normalized according to the range of motion vector changes, such as by extracting and calculating the direction and speed change values. For static regions, it is normalized based on the similarity difference of the histogram between frames. max It is the maximum value of the coefficient of variation of the data in all regions, used for dimensionless conversion, and is taken from the largest C in all regions. i′ , R i′is the data update frequency of the ith region, which is used to indicate the degree of dynamic change. For dynamic regions, it is obtained by statistically calculating the update frequency of motion pixels in consecutive frames. For static regions, it is calculated by statistically calculating the frequency of inter-frame histogram changes. max It is the maximum value of the update frequency of all regional data, indicating the reference value of frequency normalization. i′ is the storage load factor of the i′th region, which indicates the data storage occupancy of the current region. It is obtained by comprehensively considering the data volume, compression rate and current network bandwidth limitation. max is the maximum value of the storage load factor, representing the normalized base value.

[0118] For example, if there are three areas (dynamic A, static B, dynamic C), the parameters are as follows: Data change coefficient C A =90, C B =50, C C =70, the maximum value of the data variation coefficient is C max =90; Update frequency R A =20, R B =10, R C =15, the maximum update frequency is R max =20; storage load factor L A =1.5, L B =1.0, L C =1.2, the maximum storage load factor is L max =1.5.

[0119] For dynamic region A:

[0120]

[0121] For static region B:

[0122]

[0123] For dynamic region C:

[0124]

[0125] According to the calculation results, the priority is dynamic area A>dynamic area C>static area B. Dynamic area A has the highest priority because its data change coefficient and update frequency both reach the maximum value; dynamic area C has the second highest priority, and static area B has the lowest priority, which is suitable for low-frequency transmission or compression optimization. This result shows that the dimensionless formula unifies parameters of different dimensions, so that dynamic and static areas have a unified standard in priority calculation, and optimizes the allocation of transmission resources.

[0126] According to the regional priority information, the transmission capacity is allocated in real time according to the network bandwidth situation, and the camera storage image transmission data record results are obtained;

[0127] Based on the current camera cloud platform storage network bandwidth, the current transmission capacity is obtained in real time. By reading the available capacity value of the current bandwidth, it is compared with the data volume requirements of dynamic area A, dynamic area C and static area B. The motion change data and key feature points of dynamic area A are transmitted first, and a larger bandwidth is allocated according to its priority value to ensure that the dynamic change detail information is processed first. Secondly, the data content of dynamic area C is processed, and the transmission of minor change features in this area is reduced according to the bandwidth allocation conditions. Finally, for the static area B data, the transmission frequency is reduced, and only the histogram feature values ​​are extracted for transmission. When the bandwidth capacity is insufficient, the transmission of static area B is suspended first, and the transmission frequency of dynamic area C is dynamically adjusted, and the remaining bandwidth is concentrated on transmitting the key content of dynamic area A. In the case of bandwidth interruption or abnormality, the frame number and corresponding data range of the dynamic area or static area that has not completed the transmission are recorded, and finally the camera storage image transmission data record result containing transmission records, unfinished areas and untransmitted data prompts is generated.

[0128] A camera cloud platform storage method, which is performed based on the camera cloud platform storage system, includes the following steps:

[0129] S1: Based on the camera capture image, pixel brightness and color values ​​are extracted frame by frame, moving pixels are marked by inter-frame comparison, moving pixel areas are aggregated to define the boundary range, and the brightness change amplitude and frequency characteristics of non-moving areas are analyzed to mark static areas, generating dynamic and static area division results;

[0130] S2: Based on the dynamic and static area division results, the pixel brightness values ​​and color values ​​of the static area are extracted to generate histogram data, and the high similarity areas are marked by comparing the histogram similarity values ​​frame by frame. The motion vector data statistics frequency and change amplitude are extracted from the dynamic area to generate the picture repetition and dynamic feature image data;

[0131] S3: Based on the picture repetition and dynamic feature image data, the static area is classified by repetition to extract key frames or compress and store, and the dynamic area is identified and classified by motion vector frequency and change amplitude to generate optimized area storage records;

[0132] S4: Based on the optimized regional storage records, dynamically adjust the regional priority order, transmit key data and feature points to high-priority areas, give priority to transmitting dynamic area data, selectively transmit feature values ​​to low-priority static areas, and generate priority-ordered image transmission data;

[0133] S5: Based on the priority-sorted image transmission data, the transmission capacity is allocated in combination with the network bandwidth, high-priority data is transmitted first, the number and range of the untransmitted data frame are paused or recorded for the low-priority area, and the camera storage image transmission data record result is generated.

[0134] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A camera cloud platform storage system, characterized in that: The system comprises: The dynamic and static area division module compares the range of brightness and color changes based on the continuous images captured by the camera, extracts the distribution area of ​​motion pixels, determines the range of the dynamic area for the distribution area of ​​motion pixels, and identifies and marks the remaining static areas to generate dynamic and static area division results; The picture repetition evaluation module extracts pixel values ​​in the static area based on the dynamic and static area division results, obtains histogram data of the static area, calculates the similarity between the histogram data of the static area, marks the static area and repeated pictures according to the similarity, records the frequency and change value of the dynamic motion area, and generates image data of picture repetition and dynamic features; The partition storage optimization module classifies the static area according to the inter-frame repetition based on the image data of the picture repetition and dynamic characteristics, and classifies the dynamic area according to the frequency and change value of the motion vector, performs partition compression storage, and obtains optimized regional storage records; The dynamic transmission and feedback module sorts the compressed data of the dynamic and static areas according to the priority of the regional storage data based on the optimized regional storage records, optimizes the transmission of image data, and generates camera storage image transmission data record results.

2. The camera cloud platform storage system according to claim 1, characterized in that: The steps of acquiring the distribution area of ​​the motion pixels are specifically as follows: Based on the continuous picture data collected by the camera, the brightness and color values ​​of the image are extracted according to each pixel of the image frame, and the brightness and color changes of each pixel in the continuous frames are compared and analyzed to obtain the change amplitude and direction information of the pixel points; Based on the change amplitude and direction information of the pixel points, a set of pixel points whose changes exceed a threshold is determined, the area where the pixel points are located is extracted and marked, and the boundary of the dynamic area is divided to obtain the distribution area information of the moving pixels.

3. The camera cloud platform storage system according to claim 2, characterized in that: The steps for obtaining the dynamic and static area division results are specifically as follows: Based on the distribution area information of the motion pixels, densely distributed pixels are aggregated to form the dynamic area boundary by analyzing the adjacency relationship of the pixels, sparsely distributed pixels are removed, and static areas are marked by combining the brightness mean and the frequency of inter-frame changes to generate preliminary division data of dynamic and static areas; Based on the preliminary division data of the dynamic and static areas, the formula is adopted: Calculate the difference P after smoothing at the boundary between the dynamic and static areas, process the boundary between the dynamic and static areas according to the difference P, and obtain the dynamic and static area division results; Among them, n is the total number of pixels at the junction of the dynamic and static areas, B i represents the gray value of the i-th dynamic boundary pixel, S i Represents the grayscale value of the i-th static boundary pixel.

4. The camera cloud platform storage system according to claim 3, characterized in that: The specific steps for calculating the similarity between histograms are: Based on the dynamic and static area division results, the brightness values ​​and color values ​​of all pixels in the static area are extracted, the pixels are grouped and counted according to the brightness and color channels, and the number of pixels in each group is recorded to generate histogram data of the static area; For the histogram data of the static area, the formula is used: Calculate the histogram similarity value G to obtain the similarity information between the histograms; Where j represents the jth interval in the histogram, n′ is the total number of intervals in the histogram, and H Aj is the height value of the histogram of frame A in the jth interval, H Bj is the height value of the histogram of frame B in the jth interval, min(H Aj ,H Bj ) is the minimum value of the histogram height values ​​of frame A and frame B in the jth interval.

5. The camera cloud platform storage system according to claim 4, characterized in that: The steps for acquiring the image data of the picture repetitiveness and dynamic characteristics are specifically as follows: According to the similarity information between the histograms, by analyzing the similarity values ​​of the static areas frame by frame, marking the static areas above the image division threshold as high-similarity areas, marking the pictures to which the high-similarity areas belong as high-repetition pictures, and generating marking information of the high-repetition pictures; Based on the motion vector data of the moving pixels in the dynamic area, the frequency and change of the moving area are counted by analyzing the motion direction, distance and brightness change between frames, and the motion characteristics within each dynamic area are extracted, including the average frequency and change range. Combined with the marking information of the high-repetition picture, image data of the picture repetition and dynamic characteristics are obtained.

6. The camera cloud platform storage system according to claim 5, characterized in that: The steps for obtaining the optimized regional storage records are specifically as follows: Based on the image data of the picture repetition and dynamic characteristics, static areas are classified according to the inter-frame repetition, high-repetition areas and low-repetition areas are determined according to the repetition, high-repetition areas are extracted and their key frames are directly stored, and the boundary positions of the areas are recorded at the same time. For low-repetition areas, pixel values ​​are extracted frame by frame and encoded and compressed and stored in frame sequence to generate a partition storage result of repetition; Based on the motion vector data of the dynamic area, classification is performed according to the motion frequency and change value. The area with a marking frequency lower than the set frequency threshold is marked as a low-change dynamic area and its storage bit rate is reduced. For the dynamic area with a frequency higher than the frequency threshold, its motion change characteristics are extracted and combined with the position record, combined with the partition storage result of the repetitiveness, to generate an optimized regional storage record.

7. The camera cloud platform storage system according to claim 6, characterized in that: The steps for obtaining the camera storage image transmission data record result are specifically as follows: Based on the optimized regional storage records, the formula is adopted: Calculate the dimensionless priority F representing the i′th region i′ , get the regional priority information; Among them, C i′ is the data variation coefficient of the i′th region, C max is the maximum value of the coefficient of variation of data in all regions, R i′ is the data update frequency of the i′th region, R max is the maximum value of the update frequency of all regional data, L i′ is the storage load factor of the i′th region, L max is the maximum value of the storage load factor; According to the regional priority information, the transmission capacity is allocated in real time according to the network bandwidth situation, and the camera storage image transmission data record result is obtained.

8. A camera cloud platform storage method, characterized in that: The camera cloud platform storage system according to any one of claims 1 to 7 is implemented, comprising the following steps: Based on the camera capture image, pixel brightness and color values ​​are extracted frame by frame, moving pixels are marked by inter-frame comparison, moving pixel areas are aggregated to define the boundary range, and the brightness change amplitude and frequency characteristics of non-moving areas are analyzed to mark static areas, generating dynamic and static area division results; Based on the dynamic and static area division results, pixel brightness values ​​and color values ​​in the static area are extracted to generate histogram data, high similarity areas are marked by comparing the histogram similarity values ​​frame by frame, motion vector data statistics frequency and change amplitude are extracted from the dynamic area, and picture repetition and dynamic feature image data are generated; Based on the picture repetition and dynamic feature image data, extract key frames or compress and store them according to the repetition classification of static areas, identify and classify dynamic areas according to the motion vector frequency and change amplitude, and generate optimized area storage records; Based on the optimized regional storage records, dynamically adjust the regional priority order, transmit key data and feature points to high-priority areas, give priority to transmitting dynamic area data, selectively transmit feature values ​​to low-priority static areas, and generate priority-ordered image transmission data; Based on the priority-sorted image transmission data, the transmission capacity is allocated in combination with the network bandwidth, high-priority data is transmitted first, the number and range of untransmitted data frames in low-priority areas are paused or recorded, and the camera stores the image transmission data recording results.

Images