Decentralized Monitoring Method and System for Fishery-Light Complementary Photovoltaic Modules Based on Remote Sensing Images

Abstract

The present invention provides a method for monitoring decentralized fishery-photovoltaic complementary photovoltaic modules based on remote sensing images, belonging to the technical field of image recognition and intelligent monitoring. By obtaining hyperspectral images and using object detection and segmentation algorithms based on deep learning to extract the preliminary target areas of the photovoltaic modules, aiming at the problems of adhesion and overlap that may occur at the boundaries of the photovoltaic modules, such as with the water surface extension and bare land areas, by determining the overlapping and adhesive areas and according to the position results of the preliminary target areas of the plots, through the strategy of integrating decision rules for fine area segmentation, and finally obtaining the change information of the photovoltaic modules by inverting the remote sensing images, realizing high-precision real-time monitoring of the decentralized fishery-photovoltaic complementary photovoltaic modules.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and intelligent monitoring, and particularly relates to a method and system for monitoring decentralized fishery-solar complementary photovoltaic modules based on remote sensing images. Background Art

[0002] The objects of new energy power generation photovoltaic monitoring include photovoltaic arrays and distributed photovoltaics. A photovoltaic array, also known as a photovoltaic panel array, is a DC power generation unit composed of a number of photovoltaic modules assembled mechanically and electrically in a certain way and having a fixed support structure. It is mainly constructed in locations such as deserts, grasslands, cultivated lands, and water surfaces. A distributed photovoltaic power station generally refers to a power generation system that utilizes decentralized resources, has a small installed capacity, and is arranged near users. With the continuous change of the installed capacity, how to dynamically calculate the change information of all decentralized fishery-solar complementary photovoltaic modules within the monitoring range has become an urgent problem to be solved.

[0003] In the prior art, CN113298303A discloses a photovoltaic power prediction method for dynamic attention regions of meteorological satellite cloud images, which accesses a photovoltaic power station with real-time satellite remote sensing to meet the requirements of power station monitoring and operation and maintenance and real-time scheduling of the photovoltaic grid-connected system. CN117685929A discloses a method and terminal device for monitoring the spatial distribution information of photovoltaic panels. According to the position information, a high-spatial-resolution remote sensing image corresponding to the photovoltaic panel is determined for inspection and confirmation, and the user is instructed to perform fine interpretation of the photovoltaic panel boundary to extract accurate spatial distribution position information. However, with the construction of fishery-solar complementarity, on the one hand, the remote sensing image monitoring on the water surface will be affected by the water surface and the photovoltaic panels, and on the other hand, due to the decentralized characteristics of the fishery-solar complementary photovoltaic modules, it increases the difficulty of statistics and monitoring. In the prior art, no targeted monitoring method has been proposed for decentralized fishery-solar complementary photovoltaic modules, and the existing methods cannot meet the requirements of high accuracy and dynamic monitoring. Summary of the Invention

[0004] The object of the present invention is to solve the problems existing in the above prior art. In the first aspect, a method for monitoring decentralized fishery-solar complementary photovoltaic modules based on remote sensing images is provided. The method specifically includes the following technical contents:

[0005] Obtain the remote sensing image to be analyzed, specifically obtain a hyperspectral image with high resolution containing the target of decentralized fishery-solar complementary photovoltaic modules from a satellite through a satellite image acquisition unit; the hyperspectral image includes spectral features and texture features.

[0006] Perform a target detection and segmentation algorithm based on deep learning on the above hyperspectral image to extract the preliminary target region of the photovoltaic module;

[0007] Furthermore, in order to measure the difference between the prediction result and the true label in the present invention, the loss function in the target detection and segmentation algorithm is defined as follows:

[0008] loss = -(y·log(y')+(1 - y)·log(1 - y'));

[0009] Where y is the true label, which takes the value of 1 if the sample belongs to the positive sample, otherwise 0; y' is the probability that the predicted sample is a positive example.

[0010] Furthermore, perform fine region segmentation on the initially obtained target region of the photovoltaic module, specifically including

[0011] Regarding the problem of overlapping connections in the boundary extraction result of the photovoltaic module, such as adhesion and overlap may occur at the water surface extension and exposed land areas. According to the position result of the initially obtained target region, perform fine region segmentation through the strategy of fusing decision rules.

[0012] The fine region segmentation specifically includes:

[0013] Determination of overlapping and adhesive regions, to determine whether there are overlapping and adhesive regions in the result of the initially obtained target region of the photovoltaic module. Specifically: Calculate the area court(i) of the connected sub-region in each initially obtained target region and the area court_s of the initially obtained target region, and make the following judgment through the percentage A_per of the two, and the calculation formula is as follows:

[0014] A_per(i) = court(i) / court_s

[0015] Where court(i) is represented by the number of pixels in the connected sub-region, and court_s is represented by the number of pixels in the initially obtained target region. Thus, the percentage of each connected sub-region in the area of the initially obtained target region can be calculated. If there are multiple connected sub-regions and the area of the sub-region is greater than 30%, it is determined that there may be overlapping and adhesive regions in the initially obtained target region of the photovoltaic module.

[0016] For the case where there may be overlapping and adhesive regions, perform the following decision steps:

[0017] (1) If one of the connected sub-regions is dominant, such as the sub-region accounts for more than 80% of the area of the initially obtained target region, this indicates that there may be problems of mixed pixels or noise spots in the initial detection result, resulting in small deviations in the detection result. In this case, merge the connected sub-regions and, through shape constraints, remove the non-quadrilateral-shaped target regions based on the prior knowledge of the four-sided shape of the photovoltaic module, and then take the output target region as the extraction result of the photovoltaic module;

[0018] ② If two or more connected sub-regions are dominant, first use the dominant region as the inner boundary to perform mean filling on the connected sub-region, and keep the boundary of the initially obtained target region as the outer boundary;

[0019] For each filled connected sub-region, generate candidate points by randomly sampling N points (N > 20) from a uniform distribution;

[0020] Select β (β ∈ [1, N]) of the most uncertain points from the N candidate points. Determine the probability that each pixel belongs to the photovoltaic module classification. The above probability refers to a rough prediction probability of the true value. Count the number of points among the β points with a probability value greater than 0.5. Expand and merge the connected sub-regions where the ratio of the number to the number of the most uncertain points is greater than a certain threshold until all connected sub-regions are expanded and merged. Then, based on the prior knowledge of the four-side shape of the photovoltaic module, eliminate the expanded and merged regions;

[0021] Iterate the above decision step (2) until all possible overlapping and adhering regions have completed the decision, and finally output the target region of the photovoltaic module after fine segmentation.

[0022] Determine the change information of the photovoltaic module based on the target region of the photovoltaic module after fine segmentation. Specifically, compare the monitoring results at the current monitoring time point and the previous monitoring time point, and obtain the area information of the changed photovoltaic panel through hyperspectral image inversion, including the newly built area or the reduced area;

[0023] Calculate the change information of all decentralized fishery-solar complementary photovoltaic modules within the monitoring range, and output the component change information as the monitoring information of the fishery-solar complementary photovoltaic modules in this area;

[0024] Optionally, display the output monitoring information through a terminal display interface, and preferably display it visually through a web page on the photovoltaic power generation monitoring interface.

[0025] Furthermore, a monitoring system for decentralized fishery-solar complementary photovoltaic modules based on remote sensing images is provided, including:

[0026] An image acquisition module that acquires the remote sensing image to be analyzed. Specifically, obtain a hyperspectral image with high resolution containing the target of the decentralized fishery-solar complementary photovoltaic module from a satellite through a satellite image acquisition unit; the hyperspectral image includes spectral features and texture features.

[0027] A preliminary target region segmentation module that performs a target detection and segmentation algorithm based on deep learning on the above hyperspectral image to extract the preliminary target region of the photovoltaic module;

[0028] Furthermore, to measure the difference between the prediction result and the true label, the loss function in the target detection and segmentation algorithm is defined as follows:

[0029] loss = -(y·log(y′)+(1 - y)·log(1 - y′));

[0030] Where y is the true label, taking the value of 1 if the sample belongs to the positive sample, and 0 otherwise; y′ is the probability that the predicted sample is a positive example.

[0031] The fine region segmentation module performs fine region segmentation on the preliminary target region of the photovoltaic module extracted, specifically including:

[0032] Regarding the problem of overlapping connections in the extraction result of the photovoltaic module boundary, such as adhesion and overlap may occur at the water surface extension and exposed land areas. According to the position result of the preliminary target region, fine region segmentation is carried out through the strategy of fusing decision rules.

[0033] The fine region segmentation specifically includes:

[0034] Determination of overlapping and adhesive regions to determine whether there are overlapping and adhesive regions in the result of the preliminary target region of the photovoltaic module. Specifically: calculate the area court(i) of the connected sub-region in each preliminary target region and the area court_s of the preliminary target region, and make the following judgment through the percentage A_per of the two, and the calculation formula is as follows:

[0035] A_per(i) = court(i) / court_s

[0036] Where court(i) is represented by the number of pixels in the connected sub-region, and court_s is represented by the number of pixels in the preliminary target region, and thus the percentage of each connected sub-region in the area of the preliminary target region can be calculated. If there are multiple connected sub-regions and the area of the sub-region is greater than 30%, it is determined that there may be overlapping and adhesive regions in the preliminary target region of the photovoltaic module.

[0037] For the case where there may be overlapping and adhesive regions, the following decision steps are executed:

[0038] (1) If one of the connected sub-regions is dominant, such as the sub-region accounts for more than 80% of the area of the preliminary target region, this indicates that there may be problems of mixed pixels or noise spots in the preliminary detection result, resulting in small deviations in the detection result. In this case, merge the connected sub-regions and, through shape constraints, remove the non-quadrilateral-shaped target regions based on the prior knowledge of the four-side shape of the photovoltaic module, and then take the output target region as the extraction result of the photovoltaic module;

[0039] ② If two or more connected sub-regions are dominant, first use the dominant region as the inner boundary to perform mean filling on the connected sub-region, and keep the boundary of the preliminary target region as the outer boundary;

[0040] For each filled connected sub-region, generate candidate points by randomly sampling N points (N>20) from a uniform distribution;

[0041] Select β (β ∈ [1, N]) of the most uncertain points from N candidate points. Determine the probability that each pixel belongs to the photovoltaic module classification. The above probability refers to the rough prediction probability of the true value. Count the number of points among the β points with a probability value greater than 0.5. Expand and merge the connected sub-regions where the ratio of the number to the number of the most uncertain points is greater than a certain threshold until all connected sub-regions are expanded and merged. Then, based on the prior knowledge of the quadrilateral shape of the photovoltaic module, eliminate the expanded and merged regions;

[0042] Iterate the above decision step (2) until all possible overlapping and sticking regions are decided. Finally, output the target region of the photovoltaic module after fine segmentation.

[0043] The calculation module for the change information of the complementary photovoltaic module on water and solar power determines the change information of the photovoltaic module based on the target region of the photovoltaic module after fine segmentation. Specifically, it compares the monitoring results at the current monitoring time point and the previous monitoring time point, and inversely calculates the area information of the photovoltaic panels with changes through hyperspectral images, including the newly built area or the reduced area;

[0044] Calculate the change information of all distributed complementary photovoltaic modules on water and solar power within the monitoring range, and output the component change information as the monitoring information of the complementary photovoltaic module in this area;

[0045] Optionally, it further includes an output display module, which displays the output monitoring information through the terminal display interface, preferably visualizes it through a web page on the photovoltaic power generation monitoring interface.

[0046] Compared with the prior art, the beneficial effects of the present invention are: providing a brand-new monitoring method for distributed complementary photovoltaic modules on water and solar power based on remote sensing images. The obtained hyperspectral images are used to extract the preliminary target region of the photovoltaic module based on the object detection and segmentation algorithms of deep learning. In view of the possible adhesion and overlap situations in the photovoltaic module boundary extraction results, such as with the water surface extension and bare land, the overlapping and sticking regions are determined, and according to the position results of the preliminary target region of the plot, the fine region segmentation is carried out through the strategy of fusing decision rules. Finally, the change information of the photovoltaic module is inversely calculated from the remote sensing image to realize the high-precision real-time monitoring of the distributed complementary photovoltaic module on water and solar power. Description of the Drawings

[0047] Figure 1 It is the flow chart of the monitoring method for distributed complementary photovoltaic modules on water and solar power based on remote sensing images in this application;

[0048] Figure 2 It is the architecture diagram of the object detection and segmentation algorithm based on deep learning in this application;

[0049] Figure 3 It is the structural schematic diagram of an electronic device provided by the embodiment of this application. Detailed implementation mode

[0050] The present invention will be further described in detail below with reference to the accompanying drawings:

[0051] Example 1. Refer to the attached Figure 1 , a decentralized fishery-photovoltaic complementary photovoltaic module monitoring method based on remote sensing images is provided, and the method specifically includes the following technical contents:

[0052] Obtain the remote sensing image to be analyzed. Specifically, a hyperspectral image with high resolution containing the target of the decentralized fishery-photovoltaic complementary photovoltaic module is obtained from a satellite through a satellite image acquisition unit; the hyperspectral image contains spectral features and texture features.

[0053] Perform a target detection and segmentation algorithm based on deep learning on the above hyperspectral image to extract the preliminary target area of the photovoltaic module;

[0054] Furthermore, perform fine area segmentation on the extracted preliminary target area of the photovoltaic module, which specifically includes

[0055] Regarding the problem of overlapping connections in the extraction result of the photovoltaic module boundary, such as adhesion and overlap may occur with the water surface extension and bare land areas. According to the preliminary target area position result, perform fine area segmentation through a strategy of fusing decision rules.

[0056] The fine area segmentation specifically includes:

[0057] Determine the overlapping and adhesive area, and determine whether there is an overlapping and adhesive area in the preliminary target area result of the photovoltaic module. Specifically: calculate the area court(i) of the connected sub-region in each preliminary target area and the area court_s of the preliminary target area, and make the following judgment through the percentage A_per of the two. The calculation formula is as follows:

[0058] A_per(i) = court(i) / court_s

[0059] Where court(i) is represented by the number of pixels in the connected sub-region, and court_s is represented by the number of pixels in the preliminary target area. Thus, the percentage of each connected sub-region in the area of the preliminary target area can be calculated. If there are multiple connected sub-regions and the area of the sub-region is greater than 30%, it is determined that there may be an overlapping and adhesive area in the preliminary target area of the photovoltaic module.

[0060] For the situation where there may be an overlapping and adhesive area, perform the following decision steps:

[0061] (1) If one of the connected sub-regions is dominant, such as if the sub-region occupies more than 80% of the area of ​​the preliminary target region, this indicates that there may be mixed pixels or noise spots in the preliminary detection results, resulting in a small deviation in the detection results. In this case, the connected sub-regions are merged and the non-quadrilateral target regions are removed based on the prior knowledge of the quadrilateral shape of the photovoltaic module through shape constraints, and the target region is output as the photovoltaic module extraction result;

[0062] ② If there are two or more connected sub-regions that dominate, first use the dominant region as the inner boundary to perform mean filling on the connected sub-region, and retain the initial target region boundary as the outer boundary;

[0063] For each filled connected subregion, generate candidate points by randomly sampling N points (N>20) from a uniform distribution;

[0064] Select β (β∈[1,N]) most uncertain points from N candidate points. Determine the probability of each pixel belonging to the photovoltaic module classification. The above probability refers to the rough predicted probability of the true value. Count the number of β points with probability values ​​greater than 0.5, expand and merge the connected sub-regions whose ratio to the number of the most uncertain points is greater than a certain threshold, until all connected sub-regions are expanded and merged, and then remove the expanded and merged regions based on the prior knowledge of the quadrilateral shape of the photovoltaic module;

[0065] The above decision step (2) is iterated until all possible overlapping adhesion areas have been decided, and finally the finely segmented PV module target area is output.

[0066] Determine the change information of photovoltaic modules according to the target area of ​​photovoltaic modules after fine segmentation, and obtain the changed area information of photovoltaic panels through hyperspectral image inversion, including new area or reduced area, by comparing the monitoring results of the current monitoring time point with the monitoring results of the previous monitoring time point;

[0067] Calculate the change information of all distributed photovoltaic modules for fishery-photovoltaic hybrids within the monitoring range, and output the module change information as the monitoring information of photovoltaic modules for fishery-photovoltaic hybrids in the area;

[0068] Optionally, the output monitoring information is displayed through a terminal display interface, preferably visually displayed on a photovoltaic power generation monitoring interface through a web page.

[0069] Furthermore, in Example 2, a target detection and segmentation algorithm based on deep learning in the present invention is disclosed in detail. Based on this algorithm, a preliminary target area of ​​a photovoltaic module is extracted, which specifically includes the following technical contents:

[0070] Taking the remote sensing image to be analyzed as the input image, the encoder extracts the features of the input image through a deep convolutional neural network and obtains multiple effective feature layers. One of the effective feature layers is used to extract features through dilated convolutions with different dilation rates. After passing through 1×1 convolutions and an image pooling layer, the features are stacked and merged. Then, the merged result is compressed through 1×1 convolutions. The decoder has two inputs. One is the output of the deep convolutional neural network, and the other is the result after parallel dilated convolutions of the output of the above deep convolutional neural network. These two results are connected to the output result of the encoder for feature fusion, and then gradually restored to the same resolution as the original image through 3×3 convolutions and upsampling to capture the finer target boundaries of the photovoltaic modules, and the predicted segmentation result map is output to obtain the preliminary target area of the photovoltaic modules;

[0071] The decoder part adopts a structure composed of several upsampling modules and feature map fusion modules, which is used to restore the low-resolution semantic feature map output by the encoder to a high-resolution semantic segmentation result of the size of the original input image. The function of the upsampling module is to upsample the input low-resolution feature map to restore it to a high-resolution feature map. The function of the feature map fusion module is to fuse the feature maps of different scales in the encoder, and use the complementary information between the feature maps of different scales to improve the accuracy of the segmentation result. The feature map fusion module adopts a pooling pyramid module, which can extract features with different receptive fields from the feature maps of different scales, then splice these features, and fuse them through a convolutional layer.

[0072] Furthermore, in order to measure the difference between the prediction result and the true label, the loss function in the object detection and segmentation algorithm is defined as follows:

[0073] loss = -(y·log(y′)+(1 - y)·log(1 - y′));

[0074] Where y is the true label, taking the value of 1 if the sample belongs to the positive sample, otherwise taking the value of 0; y′ is the probability that the predicted sample is a positive example.

[0075] The collection of samples and the preprocessing of positive and negative samples in the present invention are common technical means in the art. Training the object detection and segmentation model based on the training samples is also well-known in the art, and will not be elaborated here.

[0076] In this embodiment, referring to the attached Figure 3 The electronic device shown also includes a bus 103 and a communication interface 104. The processor 101, the communication interface 104, and the memory 102 are connected through the bus 103.

[0077] Among them, the memory 102 may include high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk memory. The bus 103 may be an ISA bus, a PCI bus, an EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 3 only a bidirectional arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0078] The communication interface 104 is used to connect to at least one user terminal and other network units through a network interface, and send the encapsulated IPv4 packet or IPv4 packet to the user terminal through the network interface.

[0079] The processor 101 may be an integrated circuit chip with signal processing capabilities. In the implementation process, the steps of the above method can be completed by the integrated logic circuit in the hardware of the processor 101 or the instructions in the form of software. The above-mentioned processor 101 may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present disclosure. 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 disclosure can be directly embodied as being executed by a hardware decoding processor, or executed 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. This storage medium is located in the memory 102, and the processor 101 reads the information in the memory 102 and combines its hardware to complete the steps of the method in the foregoing embodiments.

[0080] The embodiment of the present invention also provides a storage medium, on which a computer program is stored, and when the computer program is run by a processor, it executes the steps of the method for monitoring decentralized fishery-photovoltaic complementary photovoltaic modules based on remote sensing images in the foregoing embodiments.

[0081] In several embodiments provided by the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be in electrical, mechanical, or other forms.

[0082] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0083] In addition, in each embodiment of the present invention, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0084] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0085] Finally, it should be noted that the above-described embodiments are only specific embodiments of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any person skilled in the technical field can still modify the technical solutions described in the foregoing embodiments, or can easily think of changes, or make equivalent replacements for some of the technical features; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0086] The above are only embodiments of the present application and are not used to limit the protection scope of the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application. It should be noted that similar reference numerals and letters indicate similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0087] The above is only the specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the technical field can easily think of changes or replacements within the technical scope disclosed in the present application, and all should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0088] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "including a..." does not exclude the existence of additional identical elements in the process, method, article or device including the element.

[0089] In the description of the present invention, unless otherwise specified, the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0090] Finally, it should be noted that the above technical solutions are only one implementation manner of the present invention. For those skilled in the art, based on the disclosed application methods and principles of the present invention, it is very easy to make various types of improvements or deformations, not limited to the methods described in the above specific implementation manners of the present invention. Therefore, the above-described manner is only preferred and does not have a restrictive meaning.

Claims

1. A distributed fishery-photovoltaic hybrid module monitoring method based on remote sensing images, characterized in that Include: Acquire the remote sensing image to be analyzed, specifically acquire the high-resolution hyperspectral image of the distributed fishery-photovoltaic complementary photovoltaic assembly target from the satellite through the satellite image acquisition unit; A deep learning-based target detection and segmentation algorithm is applied to the hyperspectral images to extract the preliminary target area of ​​the photovoltaic modules; Furthermore, the extracted preliminary target area of ​​the photovoltaic module is finely segmented, which specifically includes: In view of the overlapping connection problem in the boundary extraction results of photovoltaic modules, fine area segmentation is performed through the strategy of integrating decision rules according to the preliminary target area location results; fine area segmentation specifically includes: Overlapping and sticking area determination, determining whether the preliminary target area results of the photovoltaic module have overlapping and sticking areas; For situations where there may be overlapping adhesion areas, the following decision steps are performed: (1) If one of the connected sub-regions is dominant, such as if the sub-region occupies more than 80% of the area of ​​the preliminary target region, the connected sub-regions are merged and the non-quadrilateral target regions are removed based on the prior knowledge of the quadrilateral shape of the photovoltaic module through shape constraints, and the target region is output as the photovoltaic module extraction result; (2) If two or more connected sub-regions are dominant, the dominant region is first used as the inner boundary to perform mean filling on the connected sub-region, and the outer boundary retains the boundary of the preliminary target region; For each filled connected subregion, generate candidate points by randomly sampling N points from a uniform distribution; Select β most uncertain points from N candidate points, where β∈[1,N], and judge the probability of each pixel belonging to the PV module classification. The above probability refers to the rough predicted probability of the true value. Count the number of β points with probability values ​​greater than 0.5, expand and merge the connected sub-regions whose ratio to the number of the most uncertain points is greater than a certain threshold, until all connected sub-regions are expanded and merged, and then remove the expanded and merged regions based on the prior knowledge of the quadrilateral shape of the PV module; Iterate the above decision step (2) until all possible overlapping adhesion areas have completed the decision, and finally output the finely segmented photovoltaic module target area; Determine the photovoltaic module change information based on the finely segmented photovoltaic module target area; calculate the change information of all decentralized fishery-photovoltaic complementary photovoltaic modules within the monitoring range, and output the module change information as the monitoring information of fishery-photovoltaic complementary photovoltaic modules in the area to be analyzed.

2. The method according to claim 1, characterized in that: The loss function in the target detection and segmentation algorithm is defined as follows: loss=-(y·log(y′)+(1-y)·log(1-y′)); Where y is the true label. If the sample is a positive sample, the value is 1, otherwise it is 0; y′ is the probability that the predicted sample is a positive example.

3. The method according to claim 2, characterized in that: Specifically: Calculate the area of ​​the connected sub-region court(i) and the area of ​​the preliminary target region court_s in each preliminary target region, and make the following judgments based on the percentage A_per of the two. The calculation formula is as follows: A_per(i)=court(i) / court_s Where court(i) is represented by the number of pixels in the connected sub-region, and court_s is represented by the number of pixels in the preliminary target area. From this, the percentage of each connected sub-region in the preliminary target area can be calculated. If there are multiple connected sub-regions and the sub-region area is greater than 30%, it is judged that the preliminary target area of ​​the photovoltaic module may have overlapping adhesion areas.

4. The method according to claim 3, characterized in that: Specifically, based on the comparison of the monitoring results at the current monitoring point and the previous monitoring point, the information on the changed photovoltaic panel area, including newly built area or reduced area, is obtained through hyperspectral image inversion.

5. The method according to claim 1, characterized in that Also includes: The output monitoring information is displayed through the terminal display interface and visualized on the photovoltaic power generation monitoring interface through the web page.

6. The method according to claim 5, characterized in that: The initial target area of ​​PV panels extracted by deep learning-based target detection and segmentation algorithms includes the following technical contents: The remote sensing image to be analyzed is used as the input image. The encoder extracts the input image features through a deep convolutional neural network and obtains multiple effective feature layers. One of the effective feature layers is extracted through dilated convolutions with different dilation rates. After passing through 1×1 convolution and image pooling layers, the features are stacked and merged. The merged results are then compressed through 1×1 convolution. The decoder input consists of two parts, one is the output of the deep convolutional neural network, and the other is the result of the above deep convolutional neural network output after parallel hole convolution. These two results are connected to the above encoder output results for feature fusion, and then gradually restored to the same resolution as the original image through 3×3 convolution and upsampling to capture finer target boundaries of photovoltaic modules, output the predicted segmentation result map, and obtain the preliminary target area of ​​​​the photovoltaic module; The decoder part adopts a structure composed of several upsampling modules and feature map fusion modules, which is used to restore the low-resolution semantic feature map output by the encoder to a high-resolution semantic segmentation result of the original input image size; the upsampling module is used to upsample the input low-resolution feature map and restore it to a high-resolution feature map; the feature map fusion module is used to fuse feature maps of different scales in the encoder, and use the complementary information between feature maps of different scales to improve the accuracy of the segmentation result; the feature map fusion module adopts a pooling pyramid module, which can extract features with different receptive fields from feature maps of different scales, and then splice these features and fuse them through a convolutional layer.

7. The method according to claim 6, characterized in that: The change information of photovoltaic modules is determined based on the target area of ​​photovoltaic modules after fine segmentation. Specifically, the monitoring results of the current monitoring point are compared with the previous monitoring point, and the changed area information of photovoltaic panels, including new area or reduced area, is obtained through hyperspectral image inversion.

8. A distributed fishery-photovoltaic hybrid module monitoring system based on remote sensing images, characterized in that Include: An image acquisition module acquires the remote sensing image to be analyzed, and acquires a high-resolution hyperspectral image of the distributed fish-light complementary photovoltaic assembly target from a satellite through a satellite image acquisition unit; A preliminary target area segmentation module performs a deep learning-based target detection and segmentation algorithm on the hyperspectral image to extract the preliminary target area of ​​the photovoltaic module; The fine area segmentation module further performs fine area segmentation on the initial target area of ​​the extracted photovoltaic module, specifically including: In view of the overlapping connection problem in the boundary extraction results of photovoltaic modules, fine area segmentation is performed through the strategy of integrating decision rules according to the preliminary target area location results; fine area segmentation specifically includes: Overlapping and sticking area determination, determining whether the preliminary target area results of the photovoltaic module have overlapping and sticking areas; For situations where there may be overlapping adhesion areas, the following decision steps are performed: (1) If one of the connected sub-regions is dominant, such as if the sub-region occupies more than 80% of the area of ​​the preliminary target region, the connected sub-regions are merged and the non-quadrilateral target regions are removed based on the prior knowledge of the quadrilateral shape of the photovoltaic module through shape constraints, and the target region is output as the photovoltaic module extraction result; (2) If two or more connected sub-regions are dominant, the dominant region is first used as the inner boundary to perform mean filling on the connected sub-region, and the outer boundary retains the boundary of the preliminary target region; For each filled connected subregion, generate candidate points by randomly sampling N points from a uniform distribution; Select β most uncertain points from N candidate points, where β∈[1,N], and judge the probability of each pixel belonging to the PV module classification. The above probability refers to the rough predicted probability of the true value. Count the number of β points with probability values ​​greater than 0.5, expand and merge the connected sub-regions whose ratio to the number of the most uncertain points is greater than a certain threshold, until all connected sub-regions are expanded and merged, and then remove the expanded and merged regions based on the prior knowledge of the quadrilateral shape of the PV module; Iterate the above decision step (2) until all possible overlapping adhesion areas have completed the decision, and finally output the finely segmented photovoltaic module target area; The fishery-photovoltaic complementary photovoltaic component change information calculation module determines the photovoltaic component change information based on the finely segmented photovoltaic component target area; calculates the change information of all decentralized fishery-photovoltaic complementary photovoltaic components within the monitoring range, and outputs the component change information as the fishery-photovoltaic complementary photovoltaic component monitoring information in the area to be analyzed.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor can implement the method described in any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the program, when executed by a processor, can implement the method described in any one of claims 1 to 7.

Images