Method and system for quickly identifying oversized ground object image on edge side

Abstract

The invention provides a method and a system for quickly identifying an oversized ground object image on an edge side. The method comprises the following steps: carrying out image preprocessing on an obtained oversized ground object image; effective features are extracted from the preprocessed image; carrying out super-pixel segmentation on the image of which the effective features are extracted; constructing a ground feature classification and recognition model on the edge side, and recognizing the image after the super-pixel segmentation through the ground feature classification and recognition model to obtain a recognition result; and carrying out model optimization on the surface feature classification and identification model constructed on the edge side. According to the method and the device, the super-large ground object image can be quickly processed and recognized, and the recognition efficiency and the recognition accuracy are improved.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and particularly to a method and system for quickly recognizing ultra-large ground object images on the edge side. Background Art

[0002] With the rapid development of technologies such as remote sensing technology and unmanned aerial vehicle photogrammetry, the data volume of ground object images has shown an explosive growth. These ultra-large ground object images contain a vast amount of information, and their data scale often reaches the TB and PB levels. The traditional cloud-based processing method faces many problems such as transmission bandwidth and latency.

[0003] In many application scenarios, such as disaster monitoring, intelligent transportation management, and military reconnaissance, the recognition of ground object images requires a quick response. For example, in the event of a disaster, it is crucial to quickly identify information such as the damage situation of buildings and the blocked positions of roads in the affected area. Due to the time for data transmission and queuing for processing, the cloud-based processing method is difficult to meet this real-time requirement.

[0004] Traditional ground object image recognition algorithms have problems of high computational complexity and low efficiency when facing ultra-large-scale images. For example, some complex deep learning-based models, although having high recognition accuracy, are large in size and difficult to run on the limited computing and storage resources of edge devices.

[0005] As an emerging computing paradigm, edge computing pushes computing and data storage to the edge side closer to the data source (i.e., the image acquisition device). This can reduce the latency of data transmission and can utilize the computing resources of edge devices for rapid processing. Processing ultra-large ground object images on the edge side can complete the recognition of preliminary and key information near the source of data generation.

[0006] Therefore, the technical problem that urgently needs to be solved currently is how to provide a method and system for quickly recognizing ultra-large ground object images on the edge side, which can quickly process ultra-large ground object images, recognize ultra-large ground object images, and improve the recognition efficiency and accuracy. Summary of the Invention

[0007] The purpose of this application is to provide a method and system for quickly recognizing ultra-large ground object images on the edge side, which can quickly process ultra-large ground object images, recognize ultra-large ground object images, and improve the recognition efficiency and accuracy.

[0008] To achieve the above object, as the first aspect of the present application, the present application provides a method for quickly performing ultra-large ground object image recognition on the edge side. The method includes the following steps: performing image preprocessing on the acquired ultra-large ground object image; extracting effective features from the preprocessed image; performing superpixel segmentation on the image with the extracted effective features; constructing a ground object classification and recognition model on the edge side, and recognizing the image after superpixel segmentation through the ground object classification and recognition model to obtain a recognition result; optimizing the ground object classification and recognition model constructed on the edge side.

[0009] The method for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein the method further includes: calculating the density value of a certain ground object type according to the recognition result of the image after superpixel segmentation by the ground object classification and recognition model; optimizing the distribution of the certain ground object type according to the density value of the certain ground object type.

[0010] The method for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein optimizing the ground object classification and recognition model constructed on the edge side includes: obtaining a training sample data set and a validation sample data set; inputting the training sample data set into a neural network basic learning model for training to obtain a trained ground object classification and recognition model; inputting the validation sample data set into the trained ground object classification and recognition model for recognition to obtain validation evaluation data; calculating a recognition quality evaluation value of the ground object classification and recognition model according to the validation evaluation data; comparing the size of the recognition quality evaluation value of the ground object classification and recognition model with a preset quality threshold. If the recognition quality evaluation value of the ground object classification and recognition model is less than the preset quality threshold, the ground object classification and recognition model is optimized, otherwise, there is no need to optimize the ground object classification and recognition model.

[0011] The method for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein the validation evaluation data includes: the time for the ground object classification and recognition model to recognize the result of the object to be recognized, the accuracy rate of the result recognized by the ground object classification and recognition model, and the response time slot of the ground object classification and recognition model.

[0012] The method for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein the calculation formula for the recognition quality evaluation value of the ground object classification and recognition model is:

[0013]

[0014] Wherein, Yd represents the recognition quality evaluation value of the ground object classification and recognition model; Q1 represents the influence weight of the time for the ground object classification and recognition model to recognize the result of the object to be recognized; K represents the total number of objects to be recognized in the validation sample data set; Tr iIndicates the time when the ground object classification and recognition model recognizes the result of the i-th object to be recognized; KE represents the number of accurate recognition results of the ground object classification and recognition model for the objects to be recognized in the verification sample dataset; Q2 represents the influence weight of the recognition result accuracy of the ground object classification and recognition model; Q3 represents the influence weight of the response time slot of the ground object classification and recognition model; Th i Represents the response time slot when the ground object classification and recognition model finishes recognizing the i-th object to be recognized.

[0015] The method for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein the recognition results of the ground object classification and recognition model for the image after superpixel segmentation include: ground object type and ground object feature data.

[0016] The method for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein the image preprocessing of the obtained ultra-large ground object image includes: cropping and normalizing the obtained ultra-large ground object image.

[0017] As a second aspect of the present application, the present application provides a system for quickly performing ultra-large ground object image recognition on the edge side, which executes the method for quickly performing ultra-large ground object image recognition on the edge side. The system includes: a preprocessing module for performing image preprocessing on the obtained ultra-large ground object image; an extraction module for extracting effective features from the preprocessed image; a segmentation module for performing superpixel segmentation on the image with effective features extracted; an edge device for constructing a ground object classification and recognition model, and recognizing the image after superpixel segmentation through the ground object classification and recognition model to obtain recognition results; a model optimization module for optimizing the ground object classification and recognition model constructed on the edge side.

[0018] The system for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein the system further includes: a data processor for calculating the density value of a certain ground object type according to the recognition results of the ground object classification and recognition model for the image after superpixel segmentation; a ground object optimization module for optimizing the distribution of the certain ground object type according to the density value of the certain ground object type.

[0019] The system for quickly performing ultra-large ground object image recognition on the edge side as described above, wherein there are multiple edge devices, and each edge device constructs its own ground object classification and recognition model.

[0020] The beneficial effects achieved by the present application are as follows:

[0021] (1) The present application performs image preprocessing and extracts effective features on the obtained ultra-large ground object image, which can quickly process the ultra-large ground object image to recognize the ultra-large ground object image, improving the recognition efficiency and recognition accuracy.

[0022] (2) The present application compares the recognition quality evaluation value of the ground object classification and recognition model with the preset quality threshold. If the recognition quality evaluation value of the ground object classification and recognition model is less than the preset quality threshold, the ground object classification and recognition model is optimized; otherwise, there is no need to optimize the ground object classification and recognition model, thereby improving the recognition accuracy of the ground object classification and recognition model.

[0023] (3) By using the ground object classification and recognition model built in the edge device, the present application can identify ultra-large ground object images, reduce the latency of data transmission, and can quickly process using the computing resources of the edge device itself. Processing ultra-large ground object images on the edge side can complete the recognition of preliminary and key information near the source of data generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present application, and those skilled in the art can also obtain other drawings based on these drawings.

[0025] Figure 1 It is a flowchart of a method for quickly identifying ultra-large ground object images on the edge side according to an embodiment of the present application.

[0026] Figure 2 It is a schematic structural diagram of a system for quickly identifying ultra-large ground object images on the edge side according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0028] Embodiment 1

[0029] As Figure 1 shown, the present application provides a method for quickly identifying ultra-large ground object images on the edge side, and the method includes the following steps:

[0030] Step S1, perform image preprocessing on the acquired ultra-large ground object image.

[0031] Specifically, the image preprocessing of the obtained ultra-large feature images includes: operations such as cropping and normalizing the obtained ultra-large feature images to reduce the data volume and complexity, making it more suitable for subsequent processing. For example, the image is cropped into appropriate small pieces for parallel processing to improve the processing efficiency.

[0032] Step S2: Extract effective features from the preprocessed image.

[0033] Specifically, extract effective features from the preprocessed image to reduce the data dimension. Methods for extracting effective features include principal component analysis, which can project high-dimensional data into a low-dimensional space while retaining the main information; texture, shape, color and other features can also be extracted to provide richer information for feature recognition.

[0034] Step S3: Perform superpixel segmentation on the image from which effective features are extracted.

[0035] Specifically, the image is segmented into multiple superpixels to reduce the complexity of image post-processing.

[0036] Specifically, use a superpixel segmentation algorithm to perform content-sensitive superpixel segmentation on the image to extract feature features.

[0037] Step S4: Build a feature classification and recognition model on the edge side, and use the feature classification and recognition model to recognize the image after superpixel segmentation to obtain the recognition result.

[0038] Among them, the recognition results include: feature type and feature feature data. Feature types include: buildings, water areas, vegetation, fields, etc. Feature feature data includes: the area of the feature, the contour of the feature, the color of the feature, etc.

[0039] Specifically, select a machine learning or deep learning model suitable for edge-side computing, such as a lightweight convolutional neural network or a Transformer model, etc. Train and optimize the model so that it can accurately identify feature categories based on the extracted features and superpixel information.

[0040] As a specific embodiment of the present invention, the trained feature classification and recognition model is built into the edge device to recognize new ultra-large feature images. Input the image after preprocessing, feature extraction and superpixel segmentation, and send it into the feature classification and recognition model in the edge device to obtain the feature recognition result. In the disaster monitoring scenario, the damaged situation of the features in the disaster area can be quickly recognized.

[0041] This application deploys an edge device on the edge side to recognize ultra-large feature images, improving the recognition efficiency and

[0042] Step S5: Optimize the feature classification and recognition model built on the edge side.

[0043] Specifically, step S5 includes:

[0044] Step S510, obtaining a training sample data set and a validation sample data set.

[0045] Step S520, inputting the training sample data set into a neural network basic learning model for training to obtain a trained ground object classification and recognition model.

[0046] Step S530, inputting the validation sample data set into the trained ground object classification and recognition model for recognition to obtain validation evaluation data.

[0047] Specifically, the validation evaluation data includes: the time taken by the ground object classification and recognition model to recognize the result of the object to be recognized, the accuracy rate of the result recognized by the ground object classification and recognition model, and the response time slot of the ground object classification and recognition model. The response time slot is the response recognition duration of the recognition model for the next input object to be recognized after recognizing the current object to be recognized.

[0048] Step S540, calculating the recognition quality evaluation value of the ground object classification and recognition model according to the validation evaluation data.

[0049] Specifically, the calculation formula for the recognition quality evaluation value of the ground object classification and recognition model is:

[0050]

[0051] Among them, Yd represents the recognition quality evaluation value of the ground object classification and recognition model; Q1 represents the influence weight of the time taken by the ground object classification and recognition model to recognize the result of the object to be recognized; K represents the total number of objects to be recognized in the validation sample data set; Tr i represents the time taken by the ground object classification and recognition model to recognize the result of the i-th object to be recognized; KE represents the number of objects whose recognition results by the ground object classification and recognition model for the objects to be recognized in the validation sample data set are accurate; Q2 represents the influence weight of the accuracy rate of the result recognized by the ground object classification and recognition model; Q3 represents the influence weight of the response time slot of the ground object classification and recognition model; Th i represents the response time slot of the ground object classification and recognition model after recognizing the i-th object to be recognized.

[0052] Step S550, comparing the recognition quality evaluation value of the ground object classification and recognition model with a preset quality threshold. If the recognition quality evaluation value of the ground object classification and recognition model is less than the preset quality threshold, optimize the ground object classification and recognition model; otherwise, there is no need to optimize the ground object classification and recognition model.

[0053] This application compares the recognition quality evaluation value of the ground object classification and recognition model with the preset quality threshold. If the recognition quality evaluation value of the ground object classification and recognition model is less than the preset quality threshold, the ground object classification and recognition model is optimized. Otherwise, there is no need to optimize the ground object classification and recognition model, thereby improving the recognition accuracy of the ground object classification and recognition model.

[0054] Step S6: Calculate the density value of a certain ground object type according to the recognition result of the image after superpixel segmentation by the ground object classification and recognition model.

[0055] Specifically, the calculation formula for the density value of a certain ground object type is as follows:

[0056]

[0057] Among them, the a-th ground object and the b-th ground object are two adjacent ground objects of the same type.

[0058] Among them, MD represents the density value of a certain ground object type; S1 represents the weight influence factor of the number of occurrences of the current ground object type in a certain area; Sc represents the area of a certain area; Cz represents the number of occurrences of the current ground object type in a certain area; a and b are parameters; S2 represents the weight influence factor of the distance between two adjacent ground objects of this type in a certain area; represents the weight of the horizontal distance influence between ground objects; represents the weight of the vertical distance influence between ground objects; DH ab represents the minimum interval distance of the vertical boundary line between the a-th ground object and the b-th ground object among two adjacent ground objects; DQ ab represents the minimum interval distance of the horizontal boundary line between the a-th ground object and the b-th ground object among two adjacent ground objects.

[0059] Among them, the horizontal boundary line is a line drawn along the horizontal direction and tangent to the ground object boundary. The vertical boundary line is a line drawn along the vertical direction and tangent to the ground object boundary.

[0060] Step S7: Optimize the distribution of a certain ground object type according to the density value of the ground object type.

[0061] Specifically, if the density value of a certain ground object type is greater than the preset threshold, diffusion improvement is performed on this ground object type. For example, the number of ground objects of this type is reduced within the original area range. If the density value of a certain ground object type is less than the preset threshold, densification improvement is performed on this ground object type. For example, the number of ground objects of this type is increased within the original area range.

[0062] Embodiment 2

[0063] This application provides a system 100 for quickly performing ultra-large feature image recognition on the edge side, which executes the method for quickly performing ultra-large feature image recognition on the edge side. The system includes:

[0064] A preprocessing module 10 for performing image preprocessing on the acquired ultra-large feature image.

[0065] An extraction module 20 for extracting effective features from the preprocessed image.

[0066] A segmentation module 30 for performing superpixel segmentation on the image from which effective features are extracted.

[0067] An edge device 40 for constructing a feature classification and recognition model, and recognizing the image after superpixel segmentation through the feature classification and recognition model to obtain a recognition result.

[0068] A model optimization module 50 for optimizing the feature classification and recognition model constructed on the edge side.

[0069] A data processor 60 for calculating the density value of a certain feature type according to the recognition result of the image after superpixel segmentation by the feature classification and recognition model.

[0070] A feature optimization module 70 for optimizing the distribution of a certain feature type according to the density value of the feature type.

[0071] Among them, there are multiple edge devices 40, and each edge device 40 constructs its own feature classification and recognition model.

[0072] Specifically, the calculation formula for the recognition quality evaluation value of the feature classification and recognition model is:

[0073]

[0074] Among them, Yd represents the recognition quality evaluation value of the feature classification and recognition model; Q1 represents the influence weight of the time when the feature classification and recognition model recognizes the result of the object to be recognized; K represents the total number of objects to be recognized in the verification sample dataset; Tr i represents the time when the feature classification and recognition model recognizes the result of the i-th object to be recognized; KE represents the number of accurate recognition results of the objects to be recognized in the verification sample dataset by the feature classification and recognition model; Q2 represents the influence weight of the recognition result accuracy rate of the feature classification and recognition model; Q3 represents the influence weight of the response time slot of the feature classification and recognition model; Th i represents the response time slot when the feature classification and recognition model finishes recognizing the i-th object to be recognized.

[0075] Specifically, the calculation formula for the density value of a certain feature type is as follows:

[0076]

[0077] Among them, the a-th feature and the b-th feature are two adjacent features of the same type.

[0078] Among them, MD represents the density value of a certain feature type; S1 represents the weight influence factor of the number of the current feature type appearing in a certain area; Sc represents the area of a certain area; Cz represents the number of the current feature type appearing in a certain area; a and b are parameters; S2 represents the weight influence factor of the distance between two adjacent features of this type in a certain area; Represents the weight influence of the horizontal distance between features; Represents the weight influence of the vertical distance between features; DH ab Represents the minimum interval distance of the vertical boundary line between the a-th feature and the b-th feature among two adjacent features; DQ ab Represents the minimum interval distance of the horizontal boundary line between the a-th feature and the b-th feature among two adjacent features.

[0079] Among them, the horizontal boundary line is a line drawn along the horizontal direction and tangent to the feature boundary. The vertical boundary line is a line drawn along the vertical direction and tangent to the feature boundary.

[0080] This application also provides a computer storage medium. The computer storage medium stores computer instructions. When the computer instructions are called, they are used to execute the address mapping method of the large-capacity solid-state drive. The computer storage medium contains one or more program instructions, and one or more program instructions are used to be executed by a processor to perform a method for quickly recognizing ultra-large feature images on the edge side.

[0081] The disclosed embodiment of the present invention provides a computer-readable storage medium. Computer program instructions are stored in the computer-readable storage medium. When the computer program instructions run on a computer, the computer is made to execute the above-mentioned method for quickly recognizing ultra-large feature images on the edge side.

[0082] The embodiment of the present invention provides a processor for processing the above-mentioned method for quickly recognizing ultra-large feature images on the edge side.

[0083] In an embodiment of the present invention, the processor may be an integrated circuit chip with signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0084] It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed and completed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The processor reads the information in the storage medium and combines its hardware to complete the steps of the above method.

[0085] The storage medium may be a memory, for example, it may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories.

[0086] Among them, the non-volatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically Erasable PROM (EEPROM), or a flash memory. The volatile memory can be a Random Access Memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM).

[0087] The beneficial effects achieved by this application are as follows:

[0088] (1) This application performs image preprocessing on the acquired ultra-large ground object images and extracts effective features, enabling rapid processing of ultra-large ground object images for recognition, thereby improving the recognition efficiency and accuracy.

[0089] (2) This application compares the recognition quality evaluation value of the ground object classification recognition model with the preset quality threshold. If the recognition quality evaluation value of the ground object classification recognition model is less than the preset quality threshold, the ground object classification recognition model is optimized; otherwise, there is no need to optimize the ground object classification recognition model, thereby improving the recognition accuracy of the ground object classification recognition model.

[0090] (3) This application uses the ground object classification recognition model built in the edge device to recognize ultra-large ground object images, which can reduce the latency of data transmission and can utilize the computing resources of the edge device itself for rapid processing. Processing ultra-large ground object images on the edge side can complete the recognition of preliminary and key information near the source of data generation.

[0091] In the description of the present application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the described features. In the description of the present application, "a plurality of" means two or more, unless otherwise specifically defined.

[0092] In the description of the present application, the phrase "for example" is used to mean "serving as an example, illustration, or explanation". Any embodiment described as "for example" in the present application is not necessarily to be construed as more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. In the following description, details are set forth for purposes of explanation. It should be understood that the invention may be practiced without these specific details. In other instances, well-known structures and processes are not described in detail so as not to obscure the description of the invention with unnecessary details. Therefore, the invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed in the present application.

[0093] The foregoing is only for the embodiments of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the scope of the claims of the present invention.

Claims

1. A method for quickly performing ultra-large ground object image recognition on the edge side, characterized in that, The method includes the following steps: Perform image preprocessing on the obtained ultra-large feature image; Extract effective features from the preprocessed image; Perform superpixel segmentation on the image with the extracted effective features; Build a feature classification and recognition model on the edge side, and use the feature classification and recognition model to recognize the image after superpixel segmentation to obtain a recognition result; Optimize the feature classification and recognition model built on the edge side.

2. The method for quickly performing ultra-large feature image recognition on the edge side according to claim 1, wherein The method further includes: According to the recognition result of the image after superpixel segmentation by the feature classification and recognition model, calculate the density value of a certain feature type; Optimize the distribution of the feature type according to the density value of the feature type.

3. The method for quickly performing ultra-large ground object image recognition on the edge side according to claim 1, characterized in that, Optimizing the feature classification and recognition model built on the edge side includes: Obtain a training sample data set and a validation sample data set; Input the training sample data set into the neural network basic learning model for training to obtain a trained feature classification and recognition model; Input the validation sample data set into the trained feature classification and recognition model for recognition to obtain validation evaluation data; Calculate the recognition quality evaluation value of the feature classification and recognition model according to the validation evaluation data; Compare the recognition quality evaluation value of the feature classification and recognition model with a preset quality threshold. If the recognition quality evaluation value of the feature classification and recognition model is less than the preset quality threshold, optimize the feature classification and recognition model, otherwise, there is no need to optimize the feature classification and recognition model.

4. The method for quickly performing ultra-large feature image recognition on the edge side according to claim 3, wherein The validation evaluation data includes: the time taken by the feature classification and recognition model to recognize the object to be recognized, the accuracy of the result recognized by the feature classification and recognition model, and the response time slot of the feature classification and recognition model.

5. The method for quickly performing ultra-large ground object image recognition on the edge side according to claim 4, wherein The calculation formula for the recognition quality evaluation value of the feature classification and recognition model is: Among them, Yd represents the recognition quality evaluation value of the ground object classification and recognition model; Q1 represents the influence weight of the time when the ground object classification and recognition model recognizes the result of the object to be recognized; K represents the total number of objects to be recognized in the verification sample dataset; Tr i represents the time when the ground object classification and recognition model recognizes the result of the i-th object to be recognized; KE represents the number of accurate recognition results of the objects to be recognized in the verification sample dataset by the ground object classification and recognition model; Q2 represents the influence weight of the recognition result accuracy rate of the ground object classification and recognition model; Q3 represents the influence weight of the response time slot of the ground object classification and recognition model; Th i represents the response time slot when the ground object classification and recognition model finishes recognizing the i-th object to be recognized.

6. The method for quickly performing ultra-large feature image recognition on the edge side according to claim 1, characterized in that, The recognition result of the feature classification and recognition model for the image after superpixel segmentation includes: feature type and feature data.

7. The method for quickly performing ultra-large feature image recognition on the edge side according to claim 1, wherein Performing image preprocessing on the obtained ultra-large feature image includes: cropping and normalizing the obtained ultra-large feature image.

8. A system for quickly performing ultra-large ground object image recognition on the edge side, characterized in that, Execute the method according to any one of claims 1-7. The system includes: A preprocessing module for performing image preprocessing on the obtained ultra-large feature image; An extraction module for extracting effective features from the preprocessed image; A segmentation module for performing superpixel segmentation on the image with the extracted effective features; An edge device for building a feature classification and recognition model and using the feature classification and recognition model to recognize the image after superpixel segmentation to obtain a recognition result; A model optimization module for optimizing the feature classification and recognition model built on the edge side.

9. The system for quickly performing ultra-large ground object image recognition on the edge side according to claim 8, characterized in that, The system further includes: A data processor for calculating the density value of a certain feature type according to the recognition result of the image after superpixel segmentation by the feature classification and recognition model; A feature optimization module for optimizing the distribution of the feature type according to the density value of the feature type.

10. The system for quickly performing ultra-large ground object image recognition on the edge side according to claim 8, wherein There are multiple edge devices, and each edge device builds its own feature classification and recognition model.

Images