An image processing system based on computer vision

Abstract

The present invention is applicable to the field of image processing technology, and in particular to an image processing system based on computer vision, the system comprising: a cutting device, a sorting device, a pushing device and a pop-up device; the cutting device is used to obtain the image to be processed, and use a preset edge detection algorithm to locate the edge pixels, mark the center position of the image to be processed as the origin, and construct a cutting coordinate system; the image to be processed is cut into four first-generation quadrants, and the number of edge pixels in each first-generation quadrant is calculated. If the number is greater than a preset threshold, the first-generation quadrants are continued to be cut; the sorting device is used according to the pre-constructed definition rules. By constructing a data stack and a cutoff block, the present invention can dynamically adjust the number of blocks in the cloud according to the operation of the cloud, and by cooperating with two groups of processing nodes, it can fully optimize resource utilization and ensure image processing efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an image processing system based on computer vision. Background Art

[0002] Image processing based on computer vision is the process of analyzing, processing and understanding images using computer vision technology and algorithms. It involves converting image data into a form that can be understood and processed by computers, extracting features from images through various algorithms and models, and analyzing them. Image processing technology based on computer vision is widely used in fields such as autonomous driving, medical image analysis and security monitoring.

[0003] The existing image processing process is generally performed locally; however, local processing requires high computing power and storage resources, and has high hardware requirements. Especially when processing complex tasks or requiring large-scale data analysis, the efficiency of local processing will be much lower than cloud processing because local devices cannot provide sufficient processing power; therefore, "how to split the image and perform parallel processing locally and in the cloud" is the technical problem that the present invention needs to solve. Summary of the invention

[0004] The purpose of the present invention is to provide an image processing system based on computer vision to solve the problem of "how to split the image and perform parallel processing locally and in the cloud" raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A computer vision-based image processing system, the system comprising: a cutting device, a sorting device, a pushing device and a popping device;

[0007] The segmentation device is used to obtain the image to be processed, locate the edge pixels using a preset edge detection algorithm, mark the center position of the image to be processed as the origin, and construct a segmentation coordinate system; segment the image to be processed into four first-generation quadrants, calculate the number of edge pixels in each first-generation quadrant, and if the number is greater than a preset threshold, continue to segment the first-generation quadrant;

[0008] The sorting device is used to define high-frequency quadrants and low-frequency quadrants in the segmented image to be processed according to the pre-constructed definition rules, and merge the high-frequency quadrants; number the high-frequency quadrants and low-frequency quadrants, calculate the edge density, and arrange all the high-frequency quadrants and low-frequency quadrants in descending order of the edge density;

[0009] The push device is used to construct blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and push the blocks into the pre-constructed data stack in the order; the left and right processing nodes are respectively linked at the head of the data stack, wherein the left processing node is a cloud gateway and the right processing node is a local processing terminal;

[0010] The pop-up device is used to insert the cut-off block into the data stack, locate the position of the cut-off block in the data stack, and configure the influencing factors of the position, wherein the influencing factors at least include: network rate and cloud computing capacity; after the push is completed, the blocks above the cut-off block are popped into the left processing node, and the remaining blocks in the data stack are popped into the right processing node, and the left processing node and the right processing node are used to process the images to be processed in the blocks in parallel.

[0011] Furthermore, the cutting device comprises:

[0012] The quadrant construction module is used to obtain the image to be processed, locate the edge pixels using a preset edge detection algorithm, mark the center of the image to be processed as the origin, and construct a segmentation coordinate system;

[0013] The calculation module is used to divide the image to be processed into four first-generation quadrants, calculate the number of edge pixels in each first-generation quadrant, and continue to divide the first-generation quadrants if the number is greater than a preset threshold.

[0014] Furthermore, the sorting device comprises:

[0015] A definition module, used to define a high-frequency quadrant and a low-frequency quadrant in the segmented image to be processed according to a pre-built definition rule, and merge the high-frequency quadrant;

[0016] The sorting module is used to number the high-frequency quadrants and the low-frequency quadrants, calculate the edge density, and arrange all the high-frequency quadrants and the low-frequency quadrants in descending order of the edge density.

[0017] Further, the pushing device comprises:

[0018] An integration module, used for constructing number blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and pushing the number blocks into a pre-constructed data stack in the order;

[0019] The link module is used to link the left and right groups of processing nodes at the head of the data stack, respectively, wherein the left processing node is a cloud gateway and the right processing node is a local processing terminal.

[0020] Further, the pop-up device includes:

[0021] A position determination module, used for inserting a cutoff block into the data stack, locating the position of the cutoff block in the data stack, and configuring factors affecting the position;

[0022] The parallel processing module is used to pop the blocks above the cutoff block into the left processing node after the push is completed, and pop the remaining blocks in the data stack into the right processing node, and use the left processing node and the right processing node to parallelly process the images to be processed in the blocks.

[0023] Furthermore, the quadrant building module includes:

[0024] A selection unit, used for randomly selecting an initial point in the image to be processed, finding four edge corners of the image to be processed, and calculating the distance from the initial point to each edge corner respectively;

[0025] An offset unit is used to offset the initial point to an origin according to the distance, wherein the distance from the origin to each edge corner is the same.

[0026] Furthermore, the definition module includes:

[0027] The mounting unit is used to define all quadrants by using the definition rule, wherein the definition rule is: if the first-generation quadrant does not need to be further divided, it is defined as a low-frequency quadrant; otherwise, if it still needs to be divided, it is defined as a high-frequency quadrant, and the low-frequency quadrant mounted in the high-frequency quadrant is merged;

[0028] The statistical unit is used to count the number of edge pixels in each quadrant.

[0029] Furthermore, the sorting module includes:

[0030] A generating unit, configured to calculate the edge density of each quadrant by dividing the number of edge pixels by the total number of pixels, and generate a label consisting of the edge density;

[0031] The corresponding unit is used to establish a corresponding relationship between the number block, the high-frequency quadrant, the low-frequency quadrant and the edge density, and insert the label into the number block through the corresponding relationship.

[0032] Furthermore, the parallel processing module includes:

[0033] A mapping unit is used to build a distributed computing architecture and split it into several computing nodes, and configure the mapping between the blocks in the left processing node and the computing nodes;

[0034] The uploading unit is used to establish a communication link between the distributed computing architecture and the left processing node, cross the communication link, and upload the blocks to the corresponding computing node based on the mapping.

[0035] Furthermore, the parallel processing module also includes:

[0036] A receiving unit, used for integrating output nodes at the bottom of the data stack and embedding synthesis strategies, and receiving images processed in parallel from two groups of processing nodes;

[0037] The output unit is used to synthesize the received data using a synthesis strategy and output it to a preset terminal through an output node.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] By constructing a coordinate system, the positioning accuracy of edge pixels can be enhanced, and through multiple segmentation of the quadrant, the edge information can be greatly refined and the detection accuracy of edge pixels can be improved. By defining high-frequency quadrants and low-frequency quadrants, the image can be recursively segmented according to the edge density to further improve the segmentation accuracy of edge pixels. By constructing data stacks and cutoff blocks, the number of cloud blocks can be dynamically adjusted according to the operation of the cloud. By cooperating with two groups of processing nodes, images can be processed in parallel locally and in the cloud at the same time, fully optimizing resource utilization and ensuring image processing efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A block diagram of a computer vision-based image processing system provided by an embodiment of the present invention;

[0041] Figure 2 A block diagram of the composition of a segmentation device in a computer vision-based image processing system provided in an embodiment of the present invention;

[0042] Figure 3 A block diagram of the composition of a sorting device in a computer vision-based image processing system provided by an embodiment of the present invention;

[0043] Figure 4 A block diagram of the composition of a push device in a computer vision-based image processing system provided by an embodiment of the present invention;

[0044] Figure 5 A block diagram of the composition of a pop-up device in a computer vision-based image processing system provided by an embodiment of the present invention;

[0045] Figure 6 A block diagram of the composition of a quadrant building module in a computer vision-based image processing system provided by an embodiment of the present invention;

[0046] Figure 7 A block diagram of the components of a definition module in a computer vision-based image processing system provided by an embodiment of the present invention;

[0047] Figure 8 A block diagram of the composition of a sorting module in a computer vision-based image processing system provided by an embodiment of the present invention;

[0048] Fig. 9 A block diagram of the composition of a parallel processing module in a computer vision-based image processing system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] 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.

[0050] In Example 1, Figure 1 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown. The computer vision-based image processing system 1 includes:

[0051] The segmentation device 11 is used to obtain the image to be processed, use a preset edge detection algorithm to locate edge pixels, mark the center position of the image to be processed as the origin, and construct a segmentation coordinate system; divide the image to be processed into four first-generation quadrants, calculate the number of edge pixels in each first-generation quadrant, and if the number is greater than a preset threshold, continue to segment the first-generation quadrant.

[0052] Specifically, edge detection algorithms in the prior art are used to find edge pixels in the image to be processed and mark them; the origin is determined in the image to be processed, and a coordinate system is constructed at the origin. The image to be processed is divided into four first-generation quadrants, and the number of edge pixels in each first-generation quadrant is counted. If the number of edge pixels in a first-generation quadrant is greater than a preset threshold, the division is continued to obtain four second-generation quadrants, and so on, until the number of edge pixels in the four n-th generation quadrants is less than or equal to the preset threshold; the preset threshold should be set in advance, and there is a positive correlation between the threshold value and the computing power of the local processing terminal.

[0053] The sorting device 12 is used to define high-frequency quadrants and low-frequency quadrants in the segmented image to be processed according to a pre-constructed definition rule, and merge the high-frequency quadrants; number the high-frequency quadrants and low-frequency quadrants, calculate the edge density, and arrange all the high-frequency quadrants and low-frequency quadrants in order of the edge density from large to small.

[0054] After the segmentation is completed, all quadrants are defined, and high-frequency quadrants and low-frequency quadrants are defined.

[0055] For example, after segmenting the image to be processed, the upper left first-generation quadrant, the lower left first-generation quadrant, the upper right first-generation quadrant and the lower right first-generation quadrant are obtained, and are respectively recorded as A, B, C and D, wherein the number of edge pixels of A, B and C is less than or equal to a preset threshold, and D needs to be segmented further to obtain four second-generation quadrants, which are respectively recorded as A', B', C' and D'. If the number of edge pixels in A', B' and C' is less than or equal to the preset threshold at this time, D' is continued to be segmented, and so on. In the above segmentation process, A, B and C are defined as low-frequency quadrants, and A', B', C' are high-frequency quadrants. The advantage of this is that the quadrants can be segmented recursively to gradually reduce the number of edge pixels in the quadrants.

[0056] At the same time, the edge density of each high-frequency quadrant and low-frequency quadrant is calculated, and all high-frequency quadrants and low-frequency quadrants are arranged in descending order of edge density.

[0057] The pushing device 13 is used to construct number blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and push the number blocks into the pre-constructed data stack in the order; the left and right groups of processing nodes are respectively linked at the head of the data stack, wherein the left processing node is a cloud gateway and the right processing node is a local processing terminal.

[0058] Construct number blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and push the number blocks into the data stack in the order of arrangement. Link the left processing node and the right processing node at the head of the data stack, where the left processing node is the cloud gateway and the right processing node is the local processing terminal.

[0059] The pop-up device 14 is used to insert the cut-off block into the data stack, locate the position of the cut-off block in the data stack, and configure the influencing factors of the position, wherein the influencing factors at least include: network rate and cloud computing capacity; after the push is completed, the blocks above the cut-off block are popped into the left processing node, and the remaining blocks in the data stack are popped into the right processing node, and the left processing node and the right processing node are used to process the images to be processed in the blocks in parallel.

[0060] Insert a cutoff block into the data stack. The position of the cutoff block is determined by influencing factors, including network speed and cloud computing capacity. If the network speed is lower or the cloud computing capacity is smaller, the position of the cutoff block will be higher, and the number of blocks that can be processed using the cloud will be smaller. Push the cutoff block to the corresponding position in the data stack. After all blocks are pushed in, pop the blocks above the cutoff block into the left processing node, and pop the blocks below the cutoff block into the right processing node.

[0061] In the data stack, the closer the blocks are arranged to the head, the greater their edge density. In other words, the amount of blocks uploaded to the cloud for processing is dynamically determined based on the cloud computing capacity and network rate, and these blocks are popped into the left processing node, and the remaining blocks are popped into the right processing node. The advantage of this is that it can dynamically determine the blocks that need to be uploaded to the cloud based on the real-time network conditions, optimize resource utilization, and ensure efficient and stable image processing.

[0062] In Example 2, Figure 2 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the segmentation device 11 includes:

[0063] The quadrant construction module 111 is used to obtain the image to be processed, locate edge pixels using a preset edge detection algorithm, mark the center position of the image to be processed as the origin, and construct a segmentation coordinate system.

[0064] Use edge detection algorithms such as Sobel or Canny to find edge pixels from the image to be processed; determine the origin position and use the segmentation coordinate system to segment the image to be processed.

[0065] The calculation module 112 is used to divide the image to be processed into four first-generation quadrants, calculate the number of edge pixels in each first-generation quadrant, and continue to divide the first-generation quadrants if the number is greater than a preset threshold.

[0066] After the initial segmentation, four first-generation quadrants are obtained. If the number of edge pixels in the first-generation quadrant is greater than the preset threshold, the segmentation will continue. Since the first-generation quadrants are of the same size, the number of edge pixels can also be used to characterize the edge density.

[0067] In Example 3, Figure 3 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the sorting device 12 includes:

[0068] The definition module 121 is used to define a high-frequency quadrant and a low-frequency quadrant in the segmented image to be processed according to a pre-built definition rule, and merge the high-frequency quadrant.

[0069] After the image to be processed is segmented, the high-frequency quadrant and the low-frequency quadrant are defined using the definition rules. If a first-generation quadrant does not need to be segmented further, it is the low-frequency quadrant. If necessary, the second-generation quadrant that is segmented and does not need to be segmented again is defined as the high-frequency quadrant. In other words, the second-generation quadrant consists of the high-frequency quadrant and the quadrant that needs to be segmented.

[0070] The sorting module 122 is used to number the high-frequency quadrants and the low-frequency quadrants, calculate the edge density, and arrange all the high-frequency quadrants and the low-frequency quadrants in descending order of the edge density.

[0071] First, the defined high-frequency quadrants and low-frequency quadrants need to be numbered. The specific numbering rule is not limited. It can be an S-type sorting, and the high-frequency quadrants and low-frequency quadrants in the processed image are numbered in numerical order. After the numbering is completed, the high-frequency quadrants and low-frequency quadrants are arranged in order from large to small according to the edge density.

[0072] In Example 4, Figure 4 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the push device 13 includes:

[0073] The integration module 131 is used to construct blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and push the blocks into the pre-constructed data stack in the order.

[0074] Construct blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, synchronize the to-be-processed image corresponding to the high-frequency quadrant or the low-frequency quadrant into the blocks, and push the blocks into the constructed data stack in sequence.

[0075] The linking module 132 is used to link the left and right groups of processing nodes at the head of the data stack, respectively, wherein the left processing node is a cloud gateway and the right processing node is a local processing terminal.

[0076] The left and right groups of processing nodes are linked at the head of the data stack respectively. The data stack is mainly a collection of technologies and tools for storing, managing, processing and analyzing data. In this embodiment, the data stack is mainly used to manage and coordinate data blocks. The data stack can be simply understood as a data storage queue, with the front end of the queue being the head and the end being the bottom. After all data blocks are popped out of the data stack, they will enter the left processing node or the right processing node.

[0077] In Example 5, Figure 5 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the pop-up device 14 includes:

[0078] The position determination module 141 is used to insert a cutoff block into the data stack, locate the position of the cutoff block in the data stack, and configure the influencing factors of the position.

[0079] According to the influencing factors, the position of the cutoff block in the data stack is determined.

[0080] The parallel processing module 142 is used to pop the blocks above the cutoff block into the left processing node after the push is completed, and pop the remaining blocks in the data stack into the right processing node, and use the left processing node and the right processing node to parallelly process the images to be processed in the blocks.

[0081] After all blocks are pushed into the data stack, blocks above the cutoff block are popped into the left processing node, and blocks below the cutoff block are popped into the right processing node. The left processing node and the right processing node are used to process the images to be processed in the blocks in parallel.

[0082] In Example 6, Figure 6 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the quadrant construction module 111 includes:

[0083] The selection unit 1111 is used to randomly select an initial point in the image to be processed, find out four edge corners of the image to be processed, and calculate the distance from the initial point to each edge corner respectively.

[0084] An initial point is randomly selected in the image to be processed, and the initial point is offset by calculating the distance from the initial point to the four edge corners.

[0085] An offset unit 1112 is used to offset the initial point to an origin according to the distance, wherein the distance from the origin to each edge corner is the same,

[0086] Find a point that is the same distance from the four edge corners, offset the initial point to this point, and determine this point as the origin.

[0087] In Example 7, Figure 7 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the definition module 121 includes:

[0088] The mounting unit 1211 is used to define all quadrants using the definition rules, wherein the definition rules are: define the quadrants that do not need to be further divided as low-frequency quadrants; if division is still required, define the quadrants obtained after division that do not need to be further divided as high-frequency quadrants, and merge the high-frequency quadrants.

[0089] All quadrants are defined using definition rules, and after the high-frequency quadrant and the low-frequency quadrant are determined, the high-frequency quadrants in the same quadrant are merged.

[0090] The counting unit 1212 is used to count the number of edge pixels in each quadrant.

[0091] Calculate the number of edge pixels in each quadrant, where the quadrants refer to the high-frequency quadrant and the low-frequency quadrant.

[0092] In Example 8, Figure 8 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the sorting module 122 includes:

[0093] The generating unit 1221 is used to calculate the edge density of each quadrant by dividing the number of edge pixels by the total number of pixels, and generate a label composed of the edge density.

[0094] The number of edge pixels in each quadrant is calculated, and the edge density of each quadrant is calculated by dividing the number of edge pixels by the total number of pixels, and the edge density is used to generate labels.

[0095] The corresponding unit 1222 is used to establish a corresponding relationship between the number block, the high-frequency quadrant, the low-frequency quadrant and the edge density, and insert the label into the number block through the corresponding relationship.

[0096] Each block corresponds to a high-frequency quadrant or a low-frequency quadrant, and also corresponds to an edge density. A correspondence between the three is established and inserted into the block. The advantage of this is that the quadrant and edge density corresponding to the block can be quickly found.

[0097] In Example 9, Fig. 9 The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the parallel processing module 142 includes:

[0098] The mapping unit 1421 is used to construct a distributed computing architecture and split it into several computing nodes, and configure the mapping between the blocks in the left processing node and the computing nodes.

[0099] Construct a distributed computing architecture, where the distributed computing architecture consists of several computing nodes. When a block is popped into the left processing node, a mapping between the block and the computing node in the left processing node is constructed. Using this mapping, the blocks can be processed in parallel, further improving the processing efficiency of the blocks.

[0100] The uploading unit 1422 is used to establish a communication link between the distributed computing architecture and the left processing node, cross the communication link, and upload the blocks to the corresponding computing node based on the mapping.

[0101] Build a communication link between the distributed computing architecture and the left processing node, transfer the images to be processed in several blocks to the computing nodes, and use the computing nodes for processing.

[0102] In Example 10, Fig. 9The structure block diagram of the computer vision-based image processing system provided by an embodiment of the present invention is shown, and the parallel processing module 142 includes:

[0103] The receiving unit 1423 is used to integrate the output nodes at the bottom of the data stack and embed the synthesis strategy to receive the images processed in parallel from the two groups of processing nodes.

[0104] Insert an output node at the bottom of the data stack, where the bottom of the data stack corresponds to the head, and embed a synthesis strategy in the output node. When the image to be processed is processed in the cloud or the local processing terminal, the processed data is classified into the output node and merged using the synthesis strategy. The synthesis strategy is a specific synthesis method for the processed data.

[0105] The output unit 1424 is used to synthesize the received data using a synthesis strategy and output it to a preset terminal through an output node.

[0106] The received data is merged by using a synthesis strategy and output to a preset terminal through an output node, thereby completing the computer vision processing of the image, wherein the preset terminal is not limited and can be a system administrator terminal.

[0107] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0108] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

[0109] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An image processing system based on computer vision, characterized in that: include: Slicing equipment, sorting equipment, pushing equipment and ejecting equipment; The segmentation device is used to obtain the image to be processed, locate the edge pixels using a preset edge detection algorithm, mark the origin at the center of the image to be processed, and construct a segmentation coordinate system; Divide the image to be processed into four first-generation quadrants, calculate the number of edge pixels in each first-generation quadrant, and if the number is greater than a preset threshold, continue to divide the first-generation quadrants; The sorting device is used to define high-frequency quadrants and low-frequency quadrants in the segmented image to be processed according to the pre-constructed definition rules, and merge the high-frequency quadrants; number the high-frequency quadrants and low-frequency quadrants, calculate the edge density, and arrange all the high-frequency quadrants and low-frequency quadrants in the order of the edge density from large to small; The push device is used to construct blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and push the blocks into the pre-constructed data stack in the order; the left and right processing nodes are respectively linked at the head of the data stack, wherein the left processing node is a cloud gateway and the right processing node is a local processing terminal; The ejection device is used to insert a cutoff block into the data stack, locate the position of the cutoff block in the data stack, and configure the influencing factors of the position, wherein the influencing factors include at least: network rate and cloud computing capacity; After the push is completed, the blocks above the cutoff block are popped into the left processing node, and the remaining blocks in the data stack are popped into the right processing node, and the left processing node and the right processing node are used to process the images to be processed in the blocks in parallel.

2. The computer vision-based image processing system according to claim 1, characterized in that: The cutting device comprises: The quadrant construction module is used to obtain the image to be processed, locate the edge pixels using a preset edge detection algorithm, mark the origin at the center of the image to be processed, and construct a segmentation coordinate system; The calculation module is used to divide the image to be processed into four first-generation quadrants, calculate the number of edge pixels in each first-generation quadrant, and continue to divide the first-generation quadrants if the number is greater than a preset threshold.

3. The computer vision-based image processing system according to claim 2, characterized in that: The sorting device comprises: A definition module, used to define a high-frequency quadrant and a low-frequency quadrant in the segmented image to be processed according to a pre-built definition rule, and merge the high-frequency quadrant; The sorting module is used to number the high-frequency quadrants and the low-frequency quadrants, calculate the edge density, and arrange all the high-frequency quadrants and the low-frequency quadrants in descending order of the edge density.

4. The computer vision-based image processing system according to claim 3, characterized in that: The pushing device comprises: An integration module, used for constructing number blocks corresponding to the high-frequency quadrant and the low-frequency quadrant one by one, and pushing the number blocks into a pre-constructed data stack in the order; The link module is used to link the left and right groups of processing nodes at the head of the data stack, respectively, wherein the left processing node is a cloud gateway and the right processing node is a local processing terminal.

5. The computer vision-based image processing system according to claim 4, characterized in that: The ejection device comprises: A position determination module, used for inserting a cutoff block into the data stack, locating the position of the cutoff block in the data stack, and configuring factors affecting the position; The parallel processing module is used to pop the blocks above the cutoff block into the left processing node after the push is completed, and pop the remaining blocks in the data stack into the right processing node, and use the left processing node and the right processing node to parallelly process the images to be processed in the blocks.

6. The computer vision-based image processing system according to claim 2, characterized in that: The quadrant building blocks include: A selection unit, used for randomly selecting an initial point in the image to be processed, finding four edge corners of the image to be processed, and calculating the distance from the initial point to each edge corner respectively; An offset unit is used to offset the initial point to an origin according to the distance, wherein the distance from the origin to each edge corner is the same.

7. The computer vision-based image processing system according to claim 3, characterized in that: The definition module includes: The mounting unit is used to define all quadrants by using the definition rule, wherein the definition rule is: define the quadrants that do not need to be further divided as low-frequency quadrants, and if division is still required, define the quadrants obtained after division and do not need to be further divided as high-frequency quadrants, and merge the high-frequency quadrants; The statistical unit is used to count the number of edge pixels in each quadrant.

8. The computer vision-based image processing system according to claim 7, characterized in that: The sorting module comprises: A generating unit, configured to calculate the edge density of each quadrant by dividing the number of edge pixels by the total number of pixels, and generate a label consisting of the edge density; The corresponding unit is used to establish a corresponding relationship among the number block, the high-frequency quadrant, the low-frequency quadrant and the edge density, and insert the label into the corresponding number block through the corresponding relationship.

9. The computer vision-based image processing system according to claim 5, characterized in that: The parallel processing module comprises: A mapping unit is used to build a distributed computing architecture and split it into several computing nodes, and configure the mapping between the blocks in the left processing node and the computing nodes; The uploading unit is used to establish a communication link between the distributed computing architecture and the left processing node, cross the communication link, and upload the blocks to the corresponding computing node based on the mapping.

10. The computer vision-based image processing system according to claim 5, characterized in that: The parallel processing module also includes: A receiving unit, used for integrating output nodes at the bottom of the data stack and embedding synthesis strategies, and receiving images processed in parallel from two groups of processing nodes; The output unit is used to synthesize the received data using a synthesis strategy and output it to a preset terminal through an output node.

Images