Space debris centroid positioning method and system based on lightweight super-resolution

Abstract

The invention discloses a space debris centroid positioning method and system based on lightweight super-resolution, and relates to the technical field of centroid positioning, and the method specifically comprises the steps: obtaining to-be-processed data; inputting the to-be-processed data into a deep convolutional network for calculation, respectively inputting calculation data into a first channel and a second channel for three-stage up-sampling processing, and performing mean summation on up-sampling data output by the first channel and the second channel to obtain mean data; inputting the mean value data into the deep convolutional network to obtain super-resolution data; and carrying out centroid solving on the super-resolution data to obtain positioning data. According to the invention, a lightweight super-resolution deep convolutional network Tpf-Net is designed, an up-and-down sampling dual-channel parallel fusion super-resolution network is established, a five-layer pyramid convolutional network is combined, eight-time super-resolution data is obtained through three times of up-sampling, and high-precision positioning calculation can be realized.

Description

Technical Field

[0001] The present invention relates to the field of centroid positioning technology, and more specifically to a space debris centroid positioning method and system based on lightweight super-resolution. Background Art

[0002] With the continuous growth of global space activities, the problem of space debris has become a key factor affecting the safety of spacecraft and the sustainability of space exploration. Space debris is mainly composed of fragments of failed satellites, rocket wreckage, and collisions of small objects. They orbit the earth at high speeds and pose a serious threat to normal satellites and other space assets. Once a collision occurs, it may lead to expensive space asset losses and even threaten the lives of astronauts. Therefore, effectively monitoring and managing space debris has become an important task to ensure the long-term safety of the space environment.

[0003] Traditional imaging technology is limited by resolution and it is difficult to accurately identify and locate these space debris of various sizes and shapes. Super-resolution imaging technology, through advanced image processing algorithms, can break through the resolution limitations of physical optical systems and provide clearer and more detailed images than the original collected data. This capability is crucial for accurately tracking and analyzing space debris.

[0004] However, imaging space debris presents a unique set of challenges:

[0005] Long distance: The distance between space debris and observation equipment is usually very far, resulting in a relatively small effective area of ​​target reflection.

[0006] Low reflectivity: Many debris surfaces may not have good reflective properties, resulting in low contrast in the image.

[0007] Large background interference: The space background is complex and changeable, including stars, planets and other man-made or natural celestial bodies, which may appear as interference sources.

[0008] Few effective pixels: Especially for small debris, the number of effective pixels that can be obtained under conventional imaging conditions is extremely limited, which directly affects the accuracy of centroid positioning.

[0009] Low signal-to-noise ratio: Due to the weak signal strength and the presence of a lot of noise, it becomes very difficult to achieve high-precision centroid positioning.

[0010] Therefore, how to improve super-resolution processing efficiency and imaging quality in the process of locating the centroid of space debris is an urgent problem that technicians in this field need to solve. Summary of the invention

[0011] In view of this, the present invention provides a space debris centroid positioning method and system based on lightweight super-resolution, which overcomes the above-mentioned defects.

[0012] In order to achieve the above object, the present invention adopts the following technical solution:

[0013] A space debris centroid positioning method based on lightweight super-resolution, the specific steps are:

[0014] Get the data to be processed;

[0015] Input the data to be processed into a deep convolutional network for calculation, input the calculated data into the first channel and the second channel respectively for three-level upsampling processing, and average sum the upsampled data output by the first channel and the second channel to obtain average data;

[0016] Inputting the mean data into the deep convolutional network to obtain super-resolution data;

[0017] The centroid of the super-resolution data is solved to obtain positioning data.

[0018] Optionally, the specific steps of performing three-stage upsampling processing in the first channel are: upsampling the calculated data three times in sequence to obtain first upsampling data, second upsampling data and third upsampling data respectively.

[0019] Optionally, the specific steps of performing three-level upsampling processing in the second channel are: the first level upsampling is: downsampling the calculated data to obtain downsampled data; upsampling the downsampled data to obtain fourth upsampled data, and summing the fourth upsampled data and the first upsampled data in the same dimension to obtain first mean data; the second level upsampling is: upsampling the first mean data to obtain fifth upsampled data, and summing the fifth upsampled data and the second upsampled data to obtain second mean data; the third level upsampling is: upsampling the second mean data to obtain sixth upsampled data, and summing the sixth upsampled data and the third upsampled data to obtain third mean data.

[0020] Optionally, before upsampling or downsampling, the data needs to be calculated through the deep convolutional network.

[0021] Optionally, the deep convolutional network adopts a pyramid convolutional network.

[0022] A space debris centroid positioning system based on lightweight super-resolution, comprising:

[0023] A data acquisition module, used to obtain data to be processed;

[0024] A multi-stage upsampling module, used for inputting the data to be processed into a deep convolutional network for calculation, inputting the calculated data into the first channel and the second channel respectively for three-stage upsampling processing, and summing the upsampled data output by the first channel and the second channel to obtain mean data;

[0025] A super-resolution data acquisition module, used for inputting the mean data into the deep convolutional network to obtain super-resolution data;

[0026] The position acquisition module is used to solve the centroid of the super-resolution data to obtain positioning data.

[0027] Optionally, the multi-stage upsampling module includes a first calculation unit, a first channel processing unit and a second channel processing unit;

[0028] The first computing unit is used to compute the data to be processed using the deep convolutional network to obtain the computed data;

[0029] The first channel processing unit is used to perform three-level upsampling processing on the calculated data to obtain first channel upsampling data; the first channel upsampling data includes: first upsampling data, second upsampling data and third upsampling data;

[0030] The second channel processing unit is used to perform three-level upsampling processing on the calculated data after downsampling to obtain second channel upsampled data, and to sum the first channel upsampled data with the second channel upsampled data to obtain the mean data.

[0031] Optionally, the second channel processing unit includes:

[0032] A down-sampling subunit, used for performing down-sampling processing on the calculated data to obtain down-sampled data;

[0033] A first up-sampling sub-unit is used to perform up-sampling processing on the down-sampled data to obtain fourth up-sampled data, where the fourth up-sampled data is in the same dimension as the first up-sampled data;

[0034] a first mean summing subunit, configured to perform mean summing of the fourth up-sampled data and the first up-sampled data to obtain first mean data;

[0035] A second up-sampling subunit is used to perform up-sampling processing on the first mean data to obtain fifth up-sampled data;

[0036] a second mean summing subunit, configured to sum the fifth up-sampled data and the second up-sampled data to obtain second mean data;

[0037] A third up-sampling subunit is used to perform up-sampling processing on the second mean data to obtain sixth up-sampled data;

[0038] a third mean summing subunit, configured to perform mean summing of the sixth up-sampled data and the third up-sampled data to obtain third mean data;

[0039] The computing subunit is used to perform computing on the data before upsampling or downsampling using the deep convolutional network.

[0040] Through the above technical solutions, it can be known that compared with the prior art, the present invention provides a method and system for locating the centroid of space debris based on lightweight super-resolution, designs a lightweight super-resolution deep convolutional network Tpf-Net, and establishes a dual-channel parallel fusion super-resolution network for up and down sampling, combined with a five-layer pyramid convolutional network, and obtains 8 times super-resolution data after three upsamplings, which can achieve high-precision positioning and solving. The calculation process only includes three calculations: up and down sampling, pyramid convolution, and mean summation, and the overall architecture is simple and suitable for parallelization. Each calculation process can use multiple calculation operators in parallel, or multiple convolutions in parallel, which greatly improves the data processing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0042] Figure 1 A schematic diagram of the method provided by the present invention;

[0043] Figure 2 A schematic diagram of the lightweight super-resolution deep convolutional network architecture provided by the present invention;

[0044] Figure 3 A schematic diagram of the pyramid convolution structure provided by the present invention;

[0045] FIG4(a) is a schematic diagram of low signal-to-noise ratio non-tailing data in the test data set provided by the present invention; FIG4(b) is a schematic diagram of high signal-to-noise ratio non-tailing data in the test data set provided by the present invention; FIG4(c) is a schematic diagram of low signal-to-noise ratio tailing data in the test data set provided by the present invention; FIG4(d) is a schematic diagram of high signal-to-noise ratio tailing data in the test data set provided by the present invention;

[0046] Figure 5A schematic diagram of the centroid positioning accuracy of different algorithms under different signal-to-noise ratios provided by the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The present invention discloses a method and system for locating the centroid of space debris based on lightweight super-resolution. By constructing a super-resolution network including parallel fusion of up- and down-sampling dual channels and combining it with a five-layer pyramid convolutional network, it is possible to generate 8 times super-resolution data through three up-sampling. On this basis, the centroid of the super-resolution data is solved to achieve the goal of high-precision positioning solution, such as Figure 1 As shown, the specific steps are:

[0049] Step 1: Obtain the data to be processed;

[0050] Step 2: Input the data to be processed into the deep convolutional network for calculation, input the calculated data into the first channel and the second channel respectively for three-level upsampling processing, and sum the upsampled data output by the first channel and the second channel to obtain the mean data;

[0051] Step 3: Input the mean data into the deep convolutional network to obtain super-resolution data;

[0052] Step 4: Solve the centroid of the super-resolution data to obtain positioning data.

[0053] Furthermore, the above methods are all implemented based on a lightweight super-resolution deep convolutional network (Tpf-Net), whose structure is as follows: Figure 2 As shown, it includes: deep convolutional network layer (conv), upsampling layer and cumulative summation layer. The entire Tpf-Net implementation method is roughly divided into three steps: input is no greater than Image , enter the deep convolutional network (conv) calculation, and the results are simultaneously upsampled and downsampled to obtain The first upsampled data and Downsampled data ; First upsampled data and Downsampled data Enter the deep convolutional network (conv) calculation respectively, and the mean sum of the same dimension data output by the two channels is calculated to obtain ; Then upsample, and finally upsample to When the mean summation result is The final super-resolution data is obtained through a deep convolutional network (conv) .

[0054] Furthermore, the first up-sampled data and downsampled data The calculation process is:

[0055] (1);

[0056] in, and are upsampling and downsampling functions respectively, For the image conduct Pyramid convolution calculation, in this embodiment, is 32.

[0057] In one embodiment, the specific steps of performing three-stage upsampling processing in the first channel are: upsampling the calculated data three times in sequence to obtain first upsampling data, second upsampling data and third upsampling data respectively.

[0058] In one embodiment, the specific steps of performing three-level upsampling processing in the second channel are as follows: the first level upsampling is: performing downsampling processing on the calculated data to obtain downsampled data; performing upsampling processing on the downsampled data to obtain fourth upsampled data, and performing mean summing of the fourth upsampled data and the first upsampled data in the same dimension to obtain first mean data; the second level upsampling is: performing upsampling processing on the first mean data to obtain fifth upsampled data, and performing mean summing of the fifth upsampled data and the second upsampled data to obtain second mean data; the third level upsampling is: performing upsampling processing on the second mean data to obtain sixth upsampled data, and performing mean summing of the sixth upsampled data and the third upsampled data to obtain third mean data.

[0059] In one embodiment, before upsampling or downsampling, the data needs to be calculated through a deep convolutional network, and the calculation process is:

[0060] (2);

[0061] In the formula, is the first up-sampled data; is the second up-sampled data; It is a pyramid convolutional network; is the first upsampling data of the second channel; is the downsampled data; is the third up-sampled data; is the fourth up-sampled data; is the fifth up-sampled data; is the first mean data; is the sixth up-sampled data; is the second mean data;

[0062] The mean summation process of the upsampling results of the dual channels at all levels is as follows:

[0063] (3);

[0064] In the formula, is the third mean data; Represents average value.

[0065] The super-resolution data is:

[0066] (4);

[0067] In one embodiment, the deep convolutional network adopts a pyramid convolutional network.

[0068] Furthermore, the deep convolutional network (conv) is a five-layer pyramid convolution network (Pyramid convolution, ), whose structure is as follows Figure 3 As shown, the same convolution calculation structure is used for different input pixel sizes, specifically: for input First, 16 Convolution kernel calculation, where , and get 16 Sub-image; then through 32 groups of 16 Convolution, get 32 Sub-image; then through 16 groups of 32 Convolution, get 16 Sub-graph; finally, through 16 groups Convolution obtains the convolution Image. A pyramid and inverted pyramid cascade convolutional network is used to calculate the up- and down-sampled images, so that the up- and down-sampled images can be accurately restored.

[0069] Based on the above description, it is not difficult to see that Tpf-Net is particularly suitable for digital logic acceleration computing, mainly for:

[0070] I. Small amount of calculation: The pyramid convolution process is just a five-layer convolution, and the amount of calculation is The image is convolved to achieve the highest, requiring less than Weights ( convolution), the weight of the entire network does not exceed , which is reduced by 90% compared with the traditional super-resolution network;

[0071] II. Fixed structure: The entire computing process only includes three types of calculations: up and down sampling, pyramid convolution, and mean summation. The overall architecture is simple and suitable for parallelization. Each computing process can use multiple computing operators in parallel or multiple convolutions in parallel to improve real-time performance.

[0072] III. High parallelism: The eight pyramid convolution calculations are independent of each other, and the intermediate variables only have sequential input and output, and do not affect the order of the convolution calculations within each pyramid convolution, which is conducive to improving the parallelism.

[0073] In one embodiment, the above method is implemented and verified, and the specific steps are:

[0074] Construct a data set: Use the public data of actual space debris monitoring of the ground-based wide-angle camera array (GWAC) to make a data set for super-resolution calculation, including 800 training images and 200 test image pairs. The Tpf-Net model is used to train the data, and the test data set is tested. At the same time, in order to test and compare, the improved least squares estimation algorithm, the point spread model algorithm and the iterative adaptive window algorithm are selected as the comparison algorithms of this embodiment, and the same test data set is tested. As three commonly used mainstream centroid positioning algorithms, their centroid positioning accuracy performance is representative.

[0075] Model setting: In this embodiment, three types of Tpf-Net are designed and tested according to different convolution kernel sizes. The convolution kernel version is Tpf-Net basic version. The convolution kernel version is Tpf-Net+, The convolution kernel version is Tpf-Net++. Three versions of the convolution kernel size are used to test the effect of the convolution kernel size on the network effect.

[0076] Training configuration: The dataset of 1,000 original images was processed, and the sampling and filtering method was used to obtain 8 times lower resolution data. The 800 pairs of training images were trained using the Pytorch framework, and the training platform used NVIDIA's GPU3060Ti and the processor used was I7-10700.

[0077] Comparison of experimental results of data sets: As shown in Table 1, the method is compared with the previous three types of space debris and star target centroid positioning methods, and the accuracy and computational complexity are also compared with the three commonly used super-resolution algorithms in the past, including traditional computing models and CNN-based algorithms. Traditional algorithms are mainly divided into two types: point spread function-based and interpolation filtering. Among them, EIWA has a maximum positioning accuracy of 0.0056 pixels. All super-resolution results are tested and compared with PSNR and SSIM.

[0078] Table 1 gives the quantitative comparison of Tpf-Net with other super-resolution algorithms and centroid positioning algorithms, as well as the final centroid positioning accuracy comparison.

[0079] Table 1 Comparison of centroid positioning methods and super-resolution algorithms under different signal-to-noise ratios

[0080]

[0081] The first three algorithms use traditional calculation methods to perform high-precision positioning of the centroid, so there are no PNSR and SSIM values. The least squares method is used to predict the centroid position. When the signal-to-noise ratio is low, the error is large, reaching less than 0.2 pixels. The point spread function method can reach about 0.05 pixels at a low signal-to-noise ratio, and improving the signal-to-noise ratio does not improve the positioning accuracy. The adaptive window method based on energy iteration reaches 0.02 pixels at low signal-to-noise ratios and about 0.005 at high signal-to-noise ratios.

[0082] The three middle methods use iterative and function reconstruction networks, enhanced eight-group convolutional networks, and multi-model integrated networks to achieve super-resolution functions. Among them, EOctConv and MMSR are difficult to form training convergence when the signal-to-noise ratio is low, which is closely related to the fact that their networks include multiple types of architectures. It can be seen that previous algorithms based on CNN are difficult to achieve high positioning accuracy. This is because super-resolution intelligent methods are mainly applicable to structured images or targets. For spatial debris point target information, it is difficult to obtain good results.

[0083] The three network versions designed in this embodiment have similar centroid positioning accuracy when trained at low signal-to-noise ratio, and The convolution kernel version has higher accuracy. However, considering that The weight of the convolution kernel has increased by 4 times, but the centroid positioning accuracy has increased by less than 20%. In addition, most targets in the field of space debris are in a low signal-to-noise ratio state, so the Tpf-Net version has higher performance.

[0084] Comparison of experimental results of different types of data: The 200 test image data can be divided into four categories, 50 images each: low signal-to-noise ratio without tailing, high signal-to-noise ratio without tailing, low signal-to-noise ratio with tailing, and high signal-to-noise ratio with tailing, as shown in Figure 4 (a)-Figure 4 (d). These four types of data have different distribution characteristics. At the same time, whether there is tailing will have a great impact on the accuracy of centroid positioning. Therefore, the results of the tests on these four types of data are compared and analyzed, as shown in Table 2. Among them, tailings below 4 pixels are considered to be without tailings, and the tailing length of the tailing data is distributed between 5-10 pixels. The case where the tail exceeds 10 pixels is not considered here.

[0085] Among them, the case of low signal-to-noise ratio without tailing is the same as the case of signal-to-noise ratio <6, and the test results are already in Table 1. The accuracy of the tailing data test results in Table 2 is one order of magnitude lower than that of the no-tailing results, and the positioning accuracy of low signal-to-noise ratio tailing data is even reduced to 0.1 pixels. This also shows that for low signal-to-noise ratio tailing data, the tailing causes the overall "break" of the target, making it difficult to identify the target body and perform super-resolution prediction during the super-resolution calculation process, resulting in low centroid positioning accuracy. The use of morphological preprocessing can effectively improve the accuracy of centroid positioning due to the "break" of low signal-to-noise ratio tailing, and through up and down sampling pyramid convolution, it has better results for a single image. The dividing threshold between high signal-to-noise ratio and low signal-to-noise ratio is 6.

[0086] Table 2 Comparison of centroid positioning methods for different types of test data and super-resolution algorithms based on progressive preprocessing

[0087]

[0088] In the case of tailing, Tpf-Net++ has higher centroid positioning accuracy than Tpf-Net. This is because when the tail length is 5-10, the increase in the convolution kernel size can improve the perception of the tailing data, thereby generating better super-resolution images and improving positioning accuracy.

[0089] Comparison of experimental results of data with different signal-to-noise ratios: Test the data with different signal-to-noise ratios in the dataset to obtain the comparison of centroid positioning accuracy results under different algorithms, such as Figure 5 shown.

[0090] In general, the increase in positioning accuracy of various algorithms tends to be stable when the signal-to-noise ratio is above 4. On the one hand, the algorithm has better adaptability when the signal-to-noise ratio is high for dim targets that have a greater impact on accuracy, and can simultaneously improve positioning accuracy as the signal-to-noise ratio increases; on the other hand, most algorithms are based on data with a signal-to-noise ratio below 3, which is too affected by noise, making it difficult for target information to provide help for accuracy. Among them, the least squares method is only a simple least squares estimate of the target area and its surroundings, so the overall accuracy is poor. When the signal-to-noise ratio is low, it is difficult for the point spread function method to achieve high-precision indicators. The EIWA algorithm uses the window function iteration method to obtain higher accuracy. Compared with the lightweight super-resolution network Tpf-Net designed in this paper, the accuracy is not much different at high signal-to-noise ratios, and at low signal-to-noise ratios, Tpf-Net improves the accuracy by 40% compared to EIWA.

[0091] This embodiment designs a lightweight super-resolution deep convolutional network Tpf-Net, which establishes a parallel fusion super-resolution network with up- and down-sampling dual channels, combined with a five-layer pyramid convolutional network, and obtains 8 times super-resolution data after three up-sampling. The centroid of the obtained super-resolution data is solved to achieve high-precision positioning solution. Compared with previous algorithms, the calculation process focuses on the up- and down-sampling calculations and the pyramid convolution calculations before and after sampling, and the weight size is only less than The architecture is simple, suitable for digital logic acceleration, and suitable for on-orbit lightweight computing applications. During the experiment, coverage tests were conducted on data of different types and different signal-to-noise ratios in GWAC. Compared with previous algorithms, the centroid positioning accuracy is improved by more than 40% at low signal-to-noise ratios, and by 10 times at high signal-to-noise ratios. The accuracy comparison with previous algorithms illustrates the effectiveness of Tpf-Net.

[0092] On the other hand, this embodiment further discloses a space debris centroid positioning system based on lightweight super-resolution, which applies the above disclosed method and includes:

[0093] A data acquisition module, used to obtain data to be processed;

[0094] A multi-level upsampling module is used to input the data to be processed into the deep convolutional network for calculation, input the calculated data into the first channel and the second channel respectively for three-level upsampling processing, and average sum the upsampled data output by the first channel and the second channel to obtain average data;

[0095] The super-resolution result acquisition module inputs the mean data into the deep convolutional network to obtain super-resolution data;

[0096] The position acquisition module is used to solve the centroid of the super-resolution data and obtain the positioning data.

[0097] In one embodiment, the multi-stage upsampling module includes a first computing unit, a first channel processing unit, and a second channel processing unit;

[0098] A first computing unit, configured to use a deep convolutional network to compute the data to be processed to obtain computing data;

[0099] A first channel processing unit is used to perform three-level upsampling processing on the calculated data to obtain first channel upsampling data; the first channel upsampling data includes: first upsampling data, second upsampling data and third upsampling data;

[0100] The second channel processing unit is used to perform three-level upsampling processing on the calculated data after downsampling to obtain the second channel upsampled data, and to sum the first channel upsampled data with the second channel upsampled data to obtain the mean data.

[0101] In one embodiment, the second channel processing unit includes:

[0102] A downsampling subunit, used for downsampling the calculated data to obtain downsampled data;

[0103] A first up-sampling sub-unit is used to perform up-sampling processing on the down-sampled data to obtain fourth up-sampled data, where the fourth up-sampled data and the first up-sampled data are in the same dimension;

[0104] A first mean summing subunit, configured to perform mean summing of the fourth up-sampled data and the first up-sampled data to obtain first mean data;

[0105] A second up-sampling subunit is used to perform up-sampling processing on the first mean data to obtain fifth up-sampled data;

[0106] A second mean summing subunit, configured to sum the fifth up-sampled data and the second up-sampled data to obtain second mean data;

[0107] A third up-sampling subunit is used to perform up-sampling processing on the second mean data to obtain sixth up-sampled data;

[0108] a third mean summing subunit, configured to sum the sixth up-sampled data and the third up-sampled data to obtain third mean data;

[0109] The second computing subunit is used to perform computing on the data before upsampling or downsampling using a deep convolutional network.

[0110] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0111] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A space debris centroid positioning method based on lightweight super-resolution, characterized in that: The specific steps are: Get the data to be processed; Input the data to be processed into a deep convolutional network for calculation, input the calculated data into the first channel and the second channel respectively for three-level upsampling processing, and average sum the upsampled data output by the first channel and the second channel to obtain average data; Inputting the mean data into the deep convolutional network to obtain super-resolution data; The centroid of the super-resolution data is solved to obtain positioning data.

2. The method for locating the centroid of space debris based on lightweight super-resolution according to claim 1, characterized in that: The specific steps of performing three-level upsampling processing in the first channel are: upsampling the calculated data three times in sequence to obtain first upsampling data, second upsampling data and third upsampling data respectively.

3. The method for locating the centroid of space debris based on lightweight super-resolution according to claim 2, characterized in that: The specific steps of performing three-level upsampling processing in the second channel are as follows: the first level upsampling is: performing downsampling processing on the calculated data to obtain downsampled data; performing upsampling processing on the downsampled data to obtain fourth upsampled data, and performing mean summing of the fourth upsampled data and the first upsampled data in the same dimension to obtain first mean data; the second level upsampling is: performing upsampling processing on the first mean data to obtain fifth upsampled data, and performing mean summing of the fifth upsampled data and the second upsampled data to obtain second mean data; the third level upsampling is: performing upsampling processing on the second mean data to obtain sixth upsampled data, and performing mean summing of the sixth upsampled data and the third upsampled data to obtain third mean data.

4. The method for locating the centroid of space debris based on lightweight super-resolution according to claim 3, characterized in that: Before upsampling or downsampling, the data needs to be calculated through the deep convolutional network.

5. The method for locating the centroid of space debris based on lightweight super-resolution according to claim 1, characterized in that: The deep convolutional network adopts a pyramid convolutional network.

6. A space debris centroid positioning system based on lightweight super-resolution, characterized in that: include: A data acquisition module, used to obtain data to be processed; A multi-stage upsampling module, used for inputting the data to be processed into a deep convolutional network for calculation, inputting the calculated data into the first channel and the second channel respectively for three-stage upsampling processing, and summing the upsampled data output by the first channel and the second channel to obtain mean data; A super-resolution data acquisition module, used for inputting the mean data into the deep convolutional network to obtain super-resolution data; The position acquisition module is used to solve the centroid of the super-resolution data to obtain positioning data.

7. The space debris centroid positioning system based on lightweight super-resolution according to claim 6, characterized in that: The multi-stage upsampling module includes a first calculation unit, a first channel processing unit, and a second channel processing unit; The first computing unit is used to compute the data to be processed using the deep convolutional network to obtain the computed data; The first channel processing unit is used to perform three-level upsampling processing on the calculated data to obtain first channel upsampling data; The first channel up-sampled data includes: first up-sampled data, second up-sampled data and third up-sampled data; The second channel processing unit is used to perform three-level upsampling processing on the calculated data after downsampling to obtain second channel upsampled data, and to sum the first channel upsampled data with the second channel upsampled data to obtain the mean data.

8. The space debris centroid positioning system based on lightweight super-resolution according to claim 7, characterized in that: The second channel processing unit comprises: A down-sampling subunit, used for performing down-sampling processing on the calculated data to obtain down-sampled data; A first up-sampling sub-unit is used to perform up-sampling processing on the down-sampled data to obtain fourth up-sampled data, where the fourth up-sampled data is in the same dimension as the first up-sampled data; a first mean summing subunit, configured to perform mean summing of the fourth up-sampled data and the first up-sampled data to obtain first mean data; A second up-sampling subunit is used to perform up-sampling processing on the first mean data to obtain fifth up-sampled data; a second mean summing subunit, configured to sum the fifth up-sampled data and the second up-sampled data to obtain second mean data; A third up-sampling subunit is used to perform up-sampling processing on the second mean data to obtain sixth up-sampled data; a third mean summing subunit, configured to perform mean summing of the sixth up-sampled data and the third up-sampled data to obtain third mean data; The computing subunit is used to perform computing on the data before upsampling or downsampling using the deep convolutional network.

Images