Massive remote sensing data star-ground integrated processing method and system

Abstract

The application provides a mass remote sensing data star-ground integrated processing method and system, comprising: acquiring mass remote sensing image data; a target detection processor on a satellite processes the mass remote sensing data to obtain a frame extraction image and a target slice; a ground application system generates a slice map corresponding to the target slice, and performs target detection after image difference between the slice map and the frame extraction image; the result of target detection is counted and analyzed, and the parameters in step S2 are optimized and adjusted. Under the condition of limited resources on the satellite, the application performs initial screening of the target area, slicing and downloading of the remote sensing image data, and realizes accurate detection and identification in the ground system in combination with the downloaded frame extraction image data, effectively reduces the pressure of downloading mass remote sensing data on the satellite, maximally retains the effective area of the target in the remote sensing data, and realizes the optimal efficiency of target detection.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and more specifically, to a method and system for integrated satellite-ground processing of massive remote sensing data. Background Technology

[0002] Target detection and recognition based on remote sensing images is a crucial component of intelligent applications of remote sensing data. With the rapid increase in the amount of data output from large-area remote sensing cameras, satellite data downlink channels are insufficient to meet the demand for full data transmission, making it difficult for ground-based application systems to acquire complete input data for target detection and recognition. Furthermore, under conditions of limited on-board resources, the performance of on-board target detection and processing algorithms is inferior to that of ground-based application systems.

[0003] Therefore, it is necessary to propose a method for integrated satellite-ground processing of massive remote sensing data to meet the accuracy requirements of target detection under limited downlink resources. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for integrated satellite-ground processing of massive remote sensing data.

[0005] A method for integrated satellite-ground processing of massive remote sensing data according to the present invention includes:

[0006] Step S1: Acquire massive amounts of remote sensing image data;

[0007] Step S2: The on-board target detection processor processes the massive amount of remote sensing data to obtain frame-by-frame images and target slices;

[0008] Step S3: The ground application system generates a slice map corresponding to the target slice, and performs image difference between the slice map and the extracted frame image before target detection;

[0009] Step S4: Statistically analyze the target detection results and optimize and adjust the parameters in step S2.

[0010] Preferably, step S2 includes:

[0011] Step S2.1: Extract frames from the massive remote sensing data image sequence output by the large-area array camera on the satellite according to the preset frame interval and then download them.

[0012] Step S2.2: Under the condition of limited on-board resources, the on-board target detection processor first performs target detection processing on the massive remote sensing data of the large array camera. After processing by the traditional target detection algorithm, pixels with values ​​greater than the set threshold can be considered as targets, and pixels with values ​​less than the set threshold are considered as background. In this way, suspected target areas are initially screened.

[0013] Step S2.3: Use the connected component algorithm on the satellite to generate connected components of the suspected target, determine the slice size based on the size of the connected components, slice the target from the remote sensing image and transmit it to the ground system.

[0014] Preferably, the downloaded slice data includes the slice's location, size, and slice image grayscale value information;

[0015] Step S2.3 includes determining the region of each target using a connected component algorithm, taking the circumscribed rectangle of each target's connected component, and expanding the circumscribed rectangle outward by several pixels in the four directions (up, down, left, and right) to obtain a slice.

[0016] Preferably, the difference formula for target detection after performing image difference between the slice texture and the frame-sampling image is as follows:

[0017]

[0018] In the formula, I represents the image after difference, I qp Indicates slice mapping, I B [·] indicates the downloaded frame image, and [·] indicates the operation of taking values ​​from 0 to 255 for the differential image.

[0019] Preferably, in step S4, the parameters include the frequency of on-board frame downlink, the threshold of the on-board detection algorithm, and the slice expansion size;

[0020] The optimization adjustment includes conducting traversal experiments using different parameter combinations, recording the data download rate and detection accuracy results under different parameter combination conditions, and selecting the optimal parameter combination that meets the requirements of data download rate and detection accuracy based on the application scenario requirements.

[0021] A massive remote sensing data integrated satellite-ground processing system provided by the present invention includes:

[0022] Module M1: Acquires massive amounts of remote sensing image data;

[0023] Module M2: The on-board target detection processor processes the massive amount of remote sensing data to obtain frame-by-frame images and target slices;

[0024] Module M3: The ground application system generates a slice map corresponding to the target slice, and performs image difference between the slice map and the extracted frame image before target detection;

[0025] Module M4: Statistically analyzes the target detection results and optimizes the parameters in module M2.

[0026] Preferably, the module M2 includes:

[0027] Module M2.1: Extracts frames from the massive remote sensing data image sequence output by the large-area array camera on the satellite according to the preset frame interval and then transmits them.

[0028] Module M2.2: Under the condition of limited on-board resources, the on-board target detection processor first performs target detection processing on the massive remote sensing data of the large array camera. After processing by traditional target detection algorithms, pixels with values ​​greater than the set threshold can be considered as targets, and pixels with values ​​less than the set threshold are considered as background. This is used to initially screen out suspected target areas.

[0029] Module M2.3: Uses a connected component algorithm on the satellite to generate connected components of suspected targets, determines the slice size based on the size of the connected components, slices the targets from the remote sensing image and transmits them to the ground system.

[0030] Preferably, the downloaded slice data includes the slice's location, size, and slice image grayscale value information;

[0031] The module M2.3 includes determining the region of each target through a connected component algorithm, taking the circumscribed rectangle of each target's connected component, and expanding the circumscribed rectangle outward by several pixels in the four directions of up, down, left, and right to obtain a slice.

[0032] Preferably, the difference formula for target detection after performing image difference between the slice texture and the frame-sampling image is as follows:

[0033]

[0034] In the formula, I represents the image after difference, I qp Indicates slice mapping, I B [·] indicates the downloaded frame image, and [·] indicates the operation of taking values ​​from 0 to 255 for the differential image.

[0035] Preferably, in module M4, the parameters include the frequency of on-board frame download, the threshold of the on-board detection algorithm, and the slice expansion size;

[0036] The optimization adjustment includes conducting traversal experiments using different parameter combinations, recording the data download rate and detection accuracy results under different parameter combination conditions, and selecting the optimal parameter combination that meets the requirements of data download rate and detection accuracy based on the application scenario requirements.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. This invention performs initial screening, slicing, and downlinking of target areas in remote sensing image data under limited on-board resources. Combined with the downlinked frame-slicing image data, it achieves accurate detection and identification on the ground system, effectively reducing the pressure of downlinking massive amounts of remote sensing data from the satellite, preserving the effective area of ​​the target in the remote sensing data to the greatest extent, and achieving optimal target detection efficiency.

[0039] 2. This invention can be configured with different parameters to achieve the optimal solution based on the actual application scenario. Attached Figure Description

[0040] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0041] Figure 1 This is a schematic diagram of the working method of the present invention;

[0042] Figure 2 To download the slice texture results;

[0043] Figure 3 The results are the detection results after differential processing on the ground. Detailed Implementation

[0044] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0045] This invention performs initial screening, slicing, and downlinking of remote sensing image data for target areas under limited on-board resources. Combined with the downlinked frame-by-frame image data, it enables accurate detection and identification on the ground system, effectively reducing the pressure of downlinking massive amounts of remote sensing data from the satellite.

[0046] According to the present invention, a method for integrated satellite-ground processing of massive remote sensing data is provided, such as... Figure 1 As shown, it includes:

[0047] Step S1: Acquire massive amounts of remote sensing image data.

[0048] Step S2: The onboard target detection processor processes the massive remote sensing data to obtain frame-by-frame images and target slices. The processing includes target detection, using a connected component algorithm to generate connected components of suspected targets and determine slice sizes, slicing the targets from the remote sensing images, and transmitting them along with the frame-by-frame images to the ground system. The massive remote sensing data is from a large-area array camera. Step S2 includes:

[0049] Step S2.1: According to a preset frame interval, extract frames from the massive remote sensing image sequence output by the large-area onboard camera and download them. The downloaded extracted frames serve as the differential background image. Therefore, there is a certain time difference between the extracted frames and the current detection frame to ensure that the geometric characteristics of the target are well preserved after differential processing. Simultaneously, the time difference between the extracted frames and the current detection frame should not be too large to prevent background changes from causing numerous false alarms in the differential image, thus affecting detection performance. Based on experience, the frame interval is typically 5–20 seconds; that is, extracting one frame every 5–20 seconds is optimal. The specific frame interval depends on the camera's image generation frequency; for example, at an image generation rate of 4Hz, the interval is 20–80 frames.

[0050] Step S2.2: Under limited on-board resources, the on-board target detection processor first performs target detection processing on the massive remote sensing data from the large-area camera. After processing with traditional target detection algorithms, pixels with values ​​greater than a set threshold are considered targets, and pixels with values ​​less than the set threshold are considered background. This initial screening identifies potential target areas. Under limited on-board resources, the performance of the target detection processing method used is difficult to match that of ground-based systems. Therefore, to achieve a high detection rate for on-board target detection, it is necessary to appropriately lower the threshold of the detection method, so that the detection results contain more pixels (while also increasing the number of false alarm targets), achieving a balance between detection rate and false alarm rate.

[0051] Step S2.3: On the satellite, a connected component algorithm is used to generate connected components for suspected targets. The slice size is determined based on the size of the connected components, and the targets are sliced ​​from the remote sensing image and transmitted to the ground system. Specifically, after determining the region of each target using the connected component algorithm, the circumscribed rectangle of each target's connected component is taken. This circumscribed rectangle is then expanded outwards by a certain number of pixels in the four directions (up, down, left, and right) to obtain the slice. The slice size is related to the size of the target itself and dynamically changes with the target size. Compared to fixed-size slices, dynamic slicing can effectively reduce the amount of data transmitted while preserving the integrity of target characteristic information.

[0052] Step S3: The ground application system generates a slice texture map corresponding to the target slice, and performs target detection after image difference between the slice texture map and the downloaded frame image. That is, the ground application system generates a black background texture map for the downloaded slice, and performs target detection after difference between the black background texture map image and the downloaded frame image. Specifically, a black background image is generated on the ground based on the size of the large-area camera remote sensing image. The downloaded slice is textured in the black background image according to its position. Then, the black background texture map image and the downloaded frame image are differenced on the ground, and a more efficient detection algorithm on the ground system is used to perform accurate target detection on the differenced image to obtain the detection result. The downloaded slice data includes the slice's position, size, and slice image grayscale value information. The slice is textured onto the black background image according to its position.

[0053] The difference formula for target detection after performing image difference analysis on the slice texture and the frame-sampling image is as follows:

[0054]

[0055] In the formula, I represents the image after difference, I qp This represents a slice map, specifically a slice image with a black background. B [·] indicates the downloaded frame image, and [·] indicates the operation of taking values ​​from 0 to 255 for the differential image.

[0056] The edges of slices in a slice texture contain high-frequency image information, which can easily be misidentified as targets. By differentiating them with the downloaded frame-stripped image, edge effects can be eliminated while preserving the characteristics of the target itself.

[0057] Step S4: Statistically analyze the target detection results and optimize the parameters in steps S2.2 and S2.3. The optimized parameters include the frequency of on-board frame download, the threshold of the on-board detection algorithm, and the slice expansion size. Through a traversal method, experiments are conducted using different parameter combinations, recording the data download rate and detection accuracy under different parameter combinations. Based on the application scenario requirements of this invention, the optimal parameter combination that meets the requirements of data download rate and detection accuracy is selected to achieve optimal detection performance under low data download rate conditions.

[0058] Furthermore, in conjunction with the appendix Figure 2 and Figure 3 The method for integrated satellite-ground processing of massive remote sensing data of the present invention is described in detail below:

[0059] This embodiment is implemented based on simulated images, simulating the process of on-board image slice downloading and ground system detection. The on-board detected and downloaded image slices are as follows: Figure 2 As shown, the image after difference and further detection is as follows: Figure 3 As shown. Original image size and Figure 2 and Figure 3 The large black image is the same, with a size of 1000×1000. After slicing, the number of pixels in the five slices transmitted are 17×15, 17×17, 17×15, 17×15, and 17×17 respectively. This effectively reduces the amount of data transmitted while ensuring the integrity of the target information.

[0060] The present invention also provides a massive remote sensing data space-ground integrated processing system, which can be implemented by executing the process steps of the massive remote sensing data space-ground integrated processing method. That is, those skilled in the art can understand the massive remote sensing data space-ground integrated processing method as a preferred embodiment of the massive remote sensing data space-ground integrated processing system.

[0061] A massive remote sensing data integrated satellite-ground processing system provided by the present invention includes:

[0062] Module M1: Acquire massive amounts of remote sensing image data.

[0063] Module M2: The onboard target detection processor processes the massive remote sensing data to obtain extracted frames and target slices. Module M2 includes: Module M2.1: Extracting frames from the massive remote sensing data image sequence output by the large-area array camera on the satellite according to a preset frame interval and then downloading them. Module M2.2: Under limited onboard resources, the onboard target detection processor first performs target detection processing on the massive remote sensing data from the large-area array camera. After processing with traditional target detection algorithms, pixels with values ​​greater than a set threshold are considered targets, and pixels with values ​​less than the set threshold are considered background, thus initially filtering out suspected target areas. Module M2.3: Using a connected component algorithm on the satellite, connected components of suspected targets are generated, and the slice size is determined based on the size of the connected components. The targets are sliced ​​from the remote sensing images and downloaded to the ground system. The downloaded slice data includes the slice's location, size, and grayscale value information. The module M2.3 includes determining the region of each target through a connected component algorithm, taking the circumscribed rectangle of each target's connected component, and expanding the circumscribed rectangle outward by several pixels in the four directions of up, down, left, and right to obtain a slice.

[0064] Module M3: The ground application system generates a slice map corresponding to the target slice, and performs image difference between the slice map and the extracted frame image before target detection. The difference formula for target detection after image difference between the slice map and the extracted frame image is as follows:

[0065]

[0066] In the formula, I represents the image after difference, I qp Indicates slice mapping, I B[·] indicates the downloaded frame image, and [·] indicates the operation of taking values ​​from 0 to 255 for the differential image.

[0067] Module M4: This module statistically analyzes the target detection results and optimizes the parameters in Module M2. The parameters in Module M4 include the frequency of on-board frame download, the threshold of the on-board detection algorithm, and the slice expansion size. The optimization involves iterating through different parameter combinations, conducting experiments, recording the data download rate and detection accuracy under different parameter combinations, and selecting the optimal parameter combination that meets the requirements of both the data download rate and detection accuracy, based on the application scenario.

[0068] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, enabling the system and its various devices, modules, and units to function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered both software modules implementing the method and structures within the hardware component.

[0069] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A method for integrated satellite-ground processing of massive remote sensing data, characterized in that, include: Step S1: Acquire massive amounts of remote sensing image data; Step S2: The on-board target detection processor processes the massive amount of remote sensing data to obtain frame-by-frame images and target slices; Step S3: The ground application system generates a slice map corresponding to the target slice, and performs image difference between the slice map and the extracted frame image before target detection; Step S4: Statistically analyze the target detection results and optimize the parameters in step S2; Step S2 includes: Step S2.1: Extract frames from the massive remote sensing data image sequence output by the large-area array camera on the satellite according to the preset frame interval and then download them. Step S2.2: Under the condition of limited on-board resources, the on-board target detection processor first performs target detection processing on the massive remote sensing data of the large array camera. After processing by the traditional target detection algorithm, pixels with values ​​greater than the set threshold can be considered as targets, and pixels with values ​​less than the set threshold are considered as background. In this way, suspected target areas are initially screened. Step S2.3: Use the connected component algorithm on the satellite to generate connected components of the suspected target, determine the slice size based on the size of the connected components, slice the target from the remote sensing image and transmit it to the ground system; The downloaded slice data includes the slice's location, size, and grayscale value information of the slice image; Step S2.3 includes determining the region of each target using a connected component algorithm, taking the circumscribed rectangle of each target's connected component, and expanding the circumscribed rectangle outward by several pixels in the four directions (up, down, left, and right) to obtain a slice; performing image difference between the slice texture and the extracted frame image, and then performing the difference formula in target detection is as follows: In the formula, This represents the image after difference. Indicates slice mapping. This indicates a frame-by-frame image being downloaded. This indicates the operation of taking values ​​from 0 to 255 for the difference image.

2. The method for integrated satellite-ground processing of massive remote sensing data according to claim 1, characterized in that, In step S4, the parameters include the frequency of on-board frame download, the threshold of the on-board detection algorithm, and the slice expansion size; The optimization adjustment includes conducting traversal experiments using different parameter combinations, recording the data download rate and detection accuracy results under different parameter combination conditions, and selecting the optimal parameter combination that meets the requirements of data download rate and detection accuracy based on the application scenario requirements.

3. A massive remote sensing data integrated space-ground processing system, characterized in that, include: Module M1: Acquires massive amounts of remote sensing image data; Module M2: The on-board target detection processor processes the massive amount of remote sensing data to obtain frame-by-frame images and target slices; Module M3: The ground application system generates a slice map corresponding to the target slice, and performs image difference between the slice map and the extracted frame image before target detection; Module M4: Statistically analyzes the target detection results and optimizes the parameters in module M2; The module M2 includes: Module M2.1: Extracts frames from the massive remote sensing data image sequence output by the large-area array camera on the satellite according to the preset frame interval and then transmits them. Module M2.2: Under the condition of limited on-board resources, the on-board target detection processor first performs target detection processing on the massive remote sensing data of the large array camera. After processing by traditional target detection algorithms, pixels with values ​​greater than the set threshold can be considered as targets, and pixels with values ​​less than the set threshold are considered as background. This is used to initially screen out suspected target areas. Module M2.3: Uses a connected component algorithm on the satellite to generate connected components of suspected targets, determines the slice size based on the size of the connected components, slices the targets from the remote sensing image and transmits them to the ground system; The downloaded slice data includes the slice's location, size, and grayscale value information of the slice image; The module M2.3 includes determining the region of each target through a connected component algorithm, taking the circumscribed rectangle of each target's connected component, and expanding the circumscribed rectangle outward by several pixels in the four directions of up, down, left, and right to obtain a slice. The difference formula for target detection after performing image difference analysis on the slice texture and the frame-sampling image is as follows: In the formula, This represents the image after difference. Indicates slice mapping. This indicates a frame-by-frame image being downloaded. This indicates the operation of taking values ​​from 0 to 255 for the difference image.

4. The massive remote sensing data integrated space-ground processing system according to claim 3, characterized in that, In module M4, the parameters include the frequency of on-board frame download, the threshold of the on-board detection algorithm, and the slice expansion size; The optimization adjustment includes conducting traversal experiments using different parameter combinations, recording the data download rate and detection accuracy results under different parameter combination conditions, and selecting the optimal parameter combination that meets the requirements of data download rate and detection accuracy based on the application scenario requirements.

Images