An infrared image-based distributed photovoltaic power station photovoltaic array segmentation method, system, device and storage medium

Abstract

The application discloses a kind of distributed photovoltaic power station photovoltaic array segmentation method, system, equipment and storage medium based on infrared image, belongs to image segmentation technical field.Model training is carried out to the infrared image of roof distributed photovoltaic power station, and after the infrared image of the photovoltaic power station is detected, the effective area of the infrared image of the photovoltaic power station is obtained;After the effective area of the infrared image of the photovoltaic power station is carried out gray processing, it is normalized again;Four direction convolution kernels are respectively defined to the processed image, and convolution operation is carried out to the four direction convolution kernels;After the processed image is binarized, the image after binarization is preprocessed again;After the image after preprocessing is straight line detection, multiple straight line clusters are obtained, the median of multiple straight line clusters is taken as detection result, and the detection result is subjected to distance criterion and angle criterion respectively, and accurate detection result is screened.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image segmentation, in particular to a distributed photovoltaic array segmentation method, system and device based on infrared images of a photovoltaic power station and a storage medium. BACKGROUND

[0002] In recent years, the photovoltaic industry has developed rapidly, and the inspection and maintenance work of photovoltaic power stations has become an integral part of the industry. In order to overcome the problems of large workload and low efficiency of traditional manual inspection, more and more enterprises have begun to use unmanned aerial vehicles equipped with infrared cameras. Unlike the human eye, infrared cameras cannot distinguish between targets and interference, which leads to the need for manual verification of the videos obtained by infrared cameras, which also has the problem of low efficiency. With the rise of artificial intelligence algorithms, people have begun to apply intelligent algorithms to photovoltaic power station inspection, and the study of "hot spot effect" has been the most popular.

[0003] Roof distributed photovoltaic power stations are composed of photovoltaic modules, supports, tracks, combiner boxes, etc. The installation environment is complex, and because some roofs are color steel tile structures, the infrared images thereof have a large amount of interference, which brings great trouble to photovoltaic hot spot detection based on infrared image processing. SUMMARY

[0004] The present application solves the problem of the existing roof distributed photovoltaic power station installation environment being complex and some roofs being color steel tile structures, which leads to a large amount of interference in the infrared images thereof, and brings great trouble to photovoltaic hot spot detection based on infrared image processing.

[0005] The photovoltaic array segmentation method based on infrared images of a distributed photovoltaic power station comprises the following steps:

[0006] Step S1, infrared images of a roof distributed photovoltaic power station are collected, and after model training on the infrared images of the photovoltaic power station, a target detection SSD algorithm is used to detect the trained infrared images of the photovoltaic power station to obtain the effective area of the infrared images of the photovoltaic power station;

[0007] Step S2, after gray scale processing of the effective area of the infrared images of the photovoltaic power station, the gray scale images are normalized;

[0008] Step S3, the normalized images are respectively defined in four directions of up, down, left and right, and convolution kernels are respectively performed on the convolution kernels in the four directions;

[0009] Step S4, after adaptive threshold binarization processing of the images processed in step S3, an image noise reduction algorithm is used to preprocess the binarized images;

[0010] In step S5, the FLD straight line detection algorithm is combined with the Hough line change to detect the straight line of the preprocessed image, a plurality of straight line clusters are obtained, the median of the plurality of straight line clusters is taken as the detection result, and the distance criterion and the angle criterion are performed on the detection result to obtain the accurate detection result.

[0011] Further, in an embodiment of the present application, the model training on the infrared image of the photovoltaic power station is performed by using a caffe deep learning architecture.

[0012] Further, in an embodiment of the present application, the convolution kernel in the up and down directions is a horizontal component boundary.

[0013] The convolution kernel in the left and right directions is a vertical component boundary.

[0014] Further, in an embodiment of the present application, the image noise reduction algorithm is a combination of open operation and closed operation.

[0015] The present application discloses a distributed photovoltaic power station photovoltaic array segmentation system based on an infrared image, which comprises the following modules:

[0016] A detection module is configured to collect the infrared image of the roof distributed photovoltaic power station, and to perform model training on the infrared image of the photovoltaic power station, and then to perform detection on the trained infrared image of the photovoltaic power station by using a target detection SSD algorithm to obtain the effective area of the infrared image of the photovoltaic power station.

[0017] A processing module is configured to perform gray scale processing on the effective area of the infrared image of the photovoltaic power station, and then to perform normalization processing on the gray scale image.

[0018] A convolution module is configured to define the convolution kernel in the up, down, left and right directions of the normalized image, and to perform convolution operation on the convolution kernel in the four directions.

[0019] A preprocessing module is configured to perform adaptive threshold binarization processing on the image processed by the convolution module, and then to perform preprocessing on the binarized image by using an image noise reduction algorithm.

[0020] A screening module is configured to perform straight line detection on the preprocessed image by using the FLD straight line detection algorithm combined with the Hough line change, to obtain a plurality of straight line clusters, to take the median of the plurality of straight line clusters as the detection result, and to perform distance criterion and angle criterion on the detection result to obtain the accurate detection result.

[0021] The electronic device disclosed by the present application comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete mutual communication through the communication bus.

[0022] a memory for storing a computer program;

[0023] a processor for implementing the method steps of any of the above methods when executing the program stored on the memory.

[0024] A computer readable storage medium according to the present application stores a computer program which, when executed by a processor, implements the method steps of any of the above methods.

[0025] The present application solves the problem that the installation environment of the existing roof distributed photovoltaic power station is complex, and since part of the roof is a color steel tile structure, the infrared image has a large amount of interference, which brings great trouble to photovoltaic hot spot detection based on infrared image processing. The specific beneficial effects include:

[0026] The photovoltaic array segmentation method based on infrared images according to the present application can realize accurate segmentation of each photovoltaic module by model training, gray processing and normalization processing on the collected infrared images of the roof distributed photovoltaic power station, and then using the FLD straight line detection algorithm combined with the Hough line change. BRIEF DESCRIPTION OF DRAWINGS

[0027] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0028] Figure 1 is a flowchart of the roof distributed photovoltaic array segmentation method based on infrared images described in the detailed description;

[0029] Figure 2 is a roof distributed photovoltaic effective area detection result graph described in the detailed description;

[0030] Figure 3 is a roof distributed photovoltaic module edge detection result graph described in the detailed description;

[0031] Figure 4 is a roof distributed photovoltaic module segmentation result graph described in the detailed description. DETAILED DESCRIPTION

[0032] The various embodiments of the present application will be described below in detail with reference to the accompanying drawings. The embodiments described by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0033] The photovoltaic array segmentation method based on infrared images according to the present embodiment includes the following steps:

[0034] Step S1, collect the infrared image of the roof distributed photovoltaic power station, train the model on the photovoltaic power station infrared image, and detect the trained photovoltaic power station infrared image by using the target detection SSD algorithm to obtain the effective area of the photovoltaic power station infrared image;

[0035] Step S2, after gray processing the effective area of the photovoltaic power station infrared image, normalize the gray image;

[0036] Step S3, define the convolution kernel of the upper, lower, left and right four directions of the normalized image respectively, and perform convolution operation on the convolution kernel of the four directions respectively;

[0037] Step S4, after adaptive threshold binarization processing the image processed in step S3, the image is preprocessed by using the image noise reduction algorithm after binarization processing;

[0038] Step S5, the FLD straight line detection algorithm is combined with the Hough line change to detect the straight line of the preprocessed image, obtain a plurality of straight line clusters, take the median value of the plurality of straight line clusters as the detection result, and perform distance criterion and angle criterion on the detection result to obtain accurate detection result.

[0039] In the embodiment, the model training on the photovoltaic power station infrared image is performed by using the caffe deep learning architecture.

[0040] In the embodiment, the convolution kernel of the upper and lower directions is the horizontal component boundary.

[0041] The convolution kernel of the left and right directions is the vertical component boundary.

[0042] In the embodiment, the image noise reduction algorithm is a combination of open operation and close operation.

[0043] The embodiment is based on the infrared image-based distributed photovoltaic power station photovoltaic array segmentation method of the application, combined with Figure 1 In order to better understand the embodiment, an actual embodiment is provided:

[0044] The infrared image-based roof distributed photovoltaic array segmentation method comprises the following steps:

[0045] S1, collect the infrared image of the roof distributed photovoltaic power station, mark the sample, train the model, apply the deep learning target detection SSD algorithm to detect the photovoltaic infrared image, and obtain the effective area of the photovoltaic power station, wherein the distributed roof infrared image is shot by the infrared camera carried by the unmanned aerial vehicle;

[0046] S2, gray processing and gray image normalization processing are performed on the effective area, so as to reduce the image operation data amount and improve the image contrast;

[0047] S3, convolution operations are respectively performed on the convolution kernels defined in the upward and downward directions and the left and right directions;

[0048] S4, binaryzation processing is performed on the above result by using an adaptive threshold, and image preprocessing is realized by using an image noise reduction algorithm;

[0049] S5, a FLD straight line detection algorithm is adopted to obtain straight line clusters in combination with Hough line changes, a median value in each straight line cluster is taken, and then a final segmentation result is obtained through distance criterion and angle criterion screening.

[0050] Photovoltaic effective area detection

[0051] The infrared image of the roof distributed photovoltaic power station has complex information and contains a large amount of background interference. There is no obvious difference between the color gradient and texture shape of part of the color steel tile structure and the photovoltaic module. The traditional image segmentation cannot accurately separate the photovoltaic effective area from the background. The embodiment adopts a caffe deep learning architecture, trains a photovoltaic effective area detection model based on a target detection SSD algorithm, calls a deep neural network in an open computer vision library OpenCV, and finally realizes accurate detection of the effective area of the roof distributed photovoltaic power station, as shown in Figure 2 .

[0052] Photovoltaic effective area preprocessing

[0053] The color information in the photovoltaic infrared image is rich, and the data amount is large during image operation. Therefore, the color image is usually converted into a gray image during image processing. During the shooting process of the infrared video material, the change of the light intensity will cause great differences in the gray distribution of the video frame image. The gray of part of the frame image is too concentrated, the image contrast is low, and the segmentation of the photovoltaic module is affected. Therefore, the gray image normalization is adopted, that is, the gray distribution in the original image is diffused to an image with the entire gray level by using the gray stretching method.

[0054] Convolution operation

[0055] The common application of the convolution operation in digital image processing is sharpening and boundary extraction. Sharpening means increasing the variance of the target pixel point and the surrounding pixel points through the convolution operation, so as to highlight the image details. The embodiment increases the gradient at the boundary by the convolution operation to achieve the purpose of edge extraction. For the horizontal direction component boundary, the convolution kernels in the upward and downward directions are defined for processing. Similarly, for the vertical direction component boundary, the convolution kernels in the left and right directions are defined for processing, as shown in Figure 3 .

[0056] Adaptive threshold binarization and image denoising

[0057] In order to highlight the convolution result, the image is subjected to adaptive threshold binarization processing. The binarization of the image helps to further reduce the image data volume. At the same time, in order to remove the redundant information in the image and retain the target without loss, a binary image morphological processing is adopted.

[0058] Erosion and expansion are commonly used means in image processing, but single erosion and expansion cannot reduce noise while maintaining the target contour. Therefore, the present embodiment adopts the combination of open operation and close operation to achieve the desired purpose. The open operation is a process of first eroding and then expanding the image, which can smooth the target contour and eliminate burrs. The close operation is a process of first expanding and then eroding the image, which can connect narrow breaks and fill holes.

[0059] Straight line detection

[0060] Straight line detection is performed by combining the FLD (Fast Line Detector) straight line detection algorithm and the Hough line transformation. The FLD straight line detection uses line features instead of traditional point features for visual target recognition. The line features are less affected by external factors such as uneven lighting and changes in viewing angle, and have stronger robustness. However, due to the limitations of the flight height of the unmanned aerial vehicle and the resolution of the infrared camera, the quality of the photovoltaic infrared image is poor, and there is still a phenomenon of missed detection when using the FLD straight line detection. Therefore, the present embodiment combines the Hough line transformation for straight line detection, and takes the median value of each group of straight line clusters as the final detection result, as shown in Figure 4 .

[0061] The solar photovoltaic module is rectangular in shape, and the horizontal (Y) and vertical (X) boundaries need to be detected separately when performing image segmentation. The distance criterion and angle criterion are applied to each group of X and Y lines to filter out the accurate detection results, thereby achieving accurate segmentation of the photovoltaic module.

[0062] The distributed photovoltaic power station photovoltaic array segmentation system based on infrared images according to the present embodiment comprises the following modules:

[0063] The detection module collects the infrared image of the roof distributed photovoltaic power station, and after model training on the infrared image of the photovoltaic power station, the trained infrared image of the photovoltaic power station is detected using the target detection SSD algorithm to obtain the effective area of the infrared image of the photovoltaic power station.

[0064] The processing module performs gray scale processing on the effective area of the infrared image of the photovoltaic power station, and then performs normalization processing on the gray scale image.

[0065] The convolution module respectively defines upper, lower, left and right four direction convolution kernels for the normalized image, and respectively performs convolution operation on the four direction convolution kernels.

[0066] The preprocessing module performs adaptive threshold binarization processing on the image processed by the convolution module, and then uses an image noise reduction algorithm to pre-process the binarized image.

[0067] The screening module uses a combination of FLD straight line detection algorithm and Hough line change to detect straight lines of the pre-processed image, obtains multiple straight line clusters, takes the median of the multiple straight line clusters as the detection result, and respectively performs distance criterion and angle criterion on the detection result to obtain accurate detection results.

[0068] The electronic device described in the embodiment includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus.

[0069] The memory is used to store a computer program.

[0070] The processor is used to execute the program stored on the memory, and realize the method steps of any of the above embodiments.

[0071] The computer readable storage medium described in the embodiment stores a computer program, and the computer program is executed by the processor to realize the method steps of any of the above embodiments.

[0072] The memory in the embodiments of the application can be a volatile memory or a nonvolatile memory, or can include both volatile and nonvolatile memory. Where the nonvolatile memory is a read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example, and not limitation, many forms of RAM are available, for example, static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM). Note that the memory described herein is intended to include, among others, these and any other memory suitable for storing the data adaptively described herein.

[0073] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through wired, such as coaxial cable, optical fiber, digital subscriber line (DSL) or wireless, such as infrared, wireless, microwave, etc. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media. The available media can be magnetic media, such as floppy disks, hard disks, magnetic tapes, optical media, such as digital video discs (DVD), or semiconductor media, such as solid state discs (SSD), etc.

[0074] In the implementation process, each step of the above method can be completed by integrated logic circuit of hardware in the processor or instruction in the form of software. The steps of the method disclosed in the embodiments of the present application can be directly embodied as hardware processor execution or combined with hardware and software modules in the processor to complete the execution. The software module can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, register or other mature storage medium in the art. The storage medium is located in the memory, and the processor reads the information in the memory and combines the hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0075] It should be noted that the processor in the embodiments of the present application can be an integrated circuit chip with a signal processing capability. In the implementation process, each step of the method embodiments can be completed by the integrated logic circuit of hardware or the instruction in the form of software in the processor. The processor mentioned above can be a general processor, a digital signal processor DSP, an application specific integrated circuit ASIC, a field programmable gate array FPGA or other programmable logic devices, a discrete gate or transistor logic device, a discrete hardware component. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as a hardware code processor for execution, or a combination of hardware and software modules in the code processor for execution. The software module can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register or other mature storage medium in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and combines the hardware to complete the steps of the above method.

[0076] The above describes in detail the method, system, device and storage medium of the present application. The principle and implementation of the present application are described in this paper. The above examples are used to help understand the method and core idea of the present application. At the same time, for those skilled in the art, according to the idea of the present application, the specific implementation and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for splitting photovoltaic arrays of a distributed photovoltaic power station based on infrared images, characterized in that, It comprises the following steps: Step S1, collect the infrared image of the roof distributed photovoltaic power station, after model training on the infrared image of the photovoltaic power station, the trained infrared image of the photovoltaic power station is detected by using target detection SSD algorithm, and the effective area of the infrared image of the photovoltaic power station is obtained; Step S2, after gray processing of the effective area of the infrared image of the photovoltaic power station, the gray image is normalized; Step S3, the normalized image is respectively defined as the convolution kernel of the upper, lower, left and right four directions, and the convolution kernel of the four directions is respectively convolved; Step S4, after adaptive threshold binarization processing of the image processed in step S3, the image after binarization processing is pretreated by using image noise reduction algorithm; Step S5, the FLD straight line detection algorithm is combined with Hough line change to detect the pretreated image, a plurality of straight line clusters are obtained, the median value of the plurality of straight line clusters is taken as the detection result, and the detection result is respectively subjected to distance criterion and angle criterion to obtain accurate detection result; The convolution kernel of the upper and lower directions is the horizontal component boundary; The convolution kernel of the left and right directions is the vertical component boundary.

2. The method according to claim 1, wherein, The model training on the infrared image of the photovoltaic power station is carried out by using caffe deep learning architecture. 3.The method of claim 1, wherein, The image noise reduction algorithm is a combination of open operation and close operation.

4. A distributed photovoltaic power station photovoltaic array segmentation system based on infrared images, characterized in that, It comprises the following modules: The detection module collects the infrared image of the roof distributed photovoltaic power station, after model training on the infrared image of the photovoltaic power station, the trained infrared image of the photovoltaic power station is detected by using target detection SSD algorithm, and the effective area of the infrared image of the photovoltaic power station is obtained; The processing module, after gray processing of the effective area of the infrared image of the photovoltaic power station, the gray image is normalized; The convolution module, the normalized image is respectively defined as the convolution kernel of the upper, lower, left and right four directions, and the convolution kernel of the four directions is respectively convolved; The preprocessing module, after adaptive threshold binarization processing of the image processed in step S3, the image after binarization processing is pretreated by using image noise reduction algorithm; The screening module, the FLD straight line detection algorithm is combined with Hough line change to detect the pretreated image, a plurality of straight line clusters are obtained, the median value of the plurality of straight line clusters is taken as the detection result, and the detection result is respectively subjected to distance criterion and angle criterion to obtain accurate detection result; The convolution kernel of the upper and lower directions is the horizontal component boundary; The convolution kernel of the left and right directions is the vertical component boundary.

5. An electronic device, comprising: It comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory complete mutual communication through the communication bus; The memory is used for storing computer programs; The processor is used for executing the programs stored on the memory, and realizes the method steps of any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, The computer program is stored in the computer readable storage medium, and the computer program is executed by the processor to realize the method steps of any one of claims 1-3.

Images