CN118521839B - Photovoltaic panel defect classification method and system based on color distribution and neural network - Google Patents

Abstract

The present invention proposes a photovoltaic panel defect classification method and system based on color distribution and neural network, which belongs to the field of image processing technology. The method collects different types of defect images on photovoltaic panels; converts the defect images from RGB color space into HSV color space, quantifies the HSV color space and constructs a color distribution statistical graph, and pre-classifies the defect images based on the distribution characteristics of the color distribution statistical graph; uses BP neural network to extract effective defect feature values from each type of defect image and constructs a corresponding input vector, and sends the input vector to the trained BP neural network. The BP neural network obtains the classification result of the defect type to which the defect image belongs according to the output value of the neuron; displays the classification result to the user in the form of an image interface, and at the same time, stores the classification result in a database for subsequent analysis and query. The present invention can realize the accurate classification of mainstream defects of photovoltaic panels.

Description

Photovoltaic panel defect classification method and system based on color distribution and neural network

Technical Field

The present invention belongs to the technical field of image processing, and in particular relates to a photovoltaic panel defect classification method and system based on color distribution and neural network.

Background Art

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

As the global demand for renewable energy continues to grow, solar photovoltaic power generation, as a clean and renewable form of energy, has become increasingly important. As the core component of the solar photovoltaic power generation system, the performance of photovoltaic panels directly affects the energy conversion efficiency and operational stability of the entire system. However, during the production, transportation and long-term use of photovoltaic panels, they may be affected by a variety of factors, resulting in various defects. These defects will not only reduce the power generation efficiency of photovoltaic panels, but may also shorten their service life and even cause safety accidents. Therefore, it is of great significance to conduct in-depth research on the classification method of photovoltaic panel defects.

At present, the defect detection and classification of photovoltaic panels are still mainly based on manual field inspections. This method that relies on manual field inspections can directly and accurately observe the defects of photovoltaic panels, but the inspection process requires a lot of manpower, material resources and time to conduct comprehensive detection and classification of the defects of photovoltaic panels, so the cost is relatively high. Moreover, manual field inspections require more content to be detected, resulting in a longer detection cycle, which may affect the normal power generation of photovoltaic panels. In addition, manual field inspections are easily affected by weather. For example, manual inspections cannot be carried out during thunderstorms, and the safety risks are relatively high. Subsequently, relevant technicians began to use intelligent algorithms such as genetic algorithms to detect and classify defects in photovoltaic panels, but such methods require a large number of samples to ensure the accuracy of classification; at the same time, intelligent algorithms have high requirements for computing power, so they are not suitable for application scenarios with complex characteristic factors such as photovoltaic panel defect detection and classification. Gradually, relevant technicians began to use image processing methods to classify the defects of photovoltaic panels, but there are still some problems. For example, the existing technology generally constructs a deep learning model to extract features from the collected photovoltaic panel images to learn the features of photovoltaic panel images with different defect types, and then classify the defects of the photovoltaic panels. However, the deep learning models used are either relatively simple and have poor learning effects, or multiple deep learning models are mechanically combined, resulting in long training time and low efficiency. Moreover, when using deep learning models to extract features from photovoltaic panel defect images, the color characteristics of the defective images themselves are generally ignored, resulting in average classification accuracy for photovoltaic panel defects.

Summary of the invention

In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a photovoltaic panel defect classification method and system based on color distribution and neural network, which can achieve accurate classification of mainstream defects of photovoltaic panels.

To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

The first aspect of the present invention provides a photovoltaic panel defect classification method based on color distribution and neural network, comprising:

Different types of defect images on photovoltaic panels are collected, and the defect types of the defect images collected on the photovoltaic panels include cracks, black blocks and hot spots.

The collected defect image is converted into a color space, and the defect image is converted from the RGB color space to the HSV color space through a color mapping method; the HSV color space is quantified, and a color distribution statistical graph is constructed, and the color distribution statistical graph is used to characterize the characteristics of color distribution in the image.

The conversion method of the defect image from the RGB color space to the HSV color space by the color mapping method is:

in, and Respectively represent the maximum and minimum values of the color components in the three channels of R channel, G channel and B channel in the RGB color space; , and Respectively represent brightness, saturation, and hue in the HSV color space.

When quantizing the HSV color space, the brightness in the HSV color space is , Saturation And color tone Quantized into 4 values, 4 values, and 16 values respectively.

The color distribution statistical diagram is expressed as:

;

in, Represents a color distribution statistical graph, Represents the ratio of the number of pixels of the i- th quantized color in the color distribution statistics graph, ; and Represents the width and height of the image respectively, represents the color quantization operation, represents a color mapping operation, represents the original defect image, It represents the i- th quantized color in the quantized color space.

The defect image is pre-classified based on the color distribution characteristics, and the number of classifications is consistent with the defect types in the collected defect image.

BP neural network is used to extract effective defect feature values from each type of defect image, including defect quantity, defect area, defect area grayscale mean and defect area aspect ratio. The extracted defect feature values are constructed into corresponding input vectors according to the structural characteristics of the BP neural network input layer; the input vectors are sent to the trained BP neural network, and the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron.

The BP neural network comprises an input layer, two hidden layers and an output layer; the number of nodes in the input layer is consistent with the number of defect feature values, and the number of nodes in the output layer is consistent with the number of defect types in the defect image.

The classification results are displayed to the user in the form of a graphical interface and stored in a database for subsequent analysis and query.

A second aspect of the present invention provides a photovoltaic panel defect classification system based on color distribution and neural network, comprising:

The image acquisition module is configured to: acquire images of different types of defects on the photovoltaic panel.

The color distribution pre-classification module is configured to: perform color space conversion on the acquired defect image, convert the defect image from the RGB color space into the representation of the HSV color space through a color mapping method; quantify the HSV color space and construct a color distribution statistical graph, and use the color distribution statistical graph to characterize the characteristics of color distribution in the image; pre-classify the defect image based on the color distribution characteristics, and the number of classifications is consistent with the type of defects in the acquired defect image.

The defect image classification module is configured as follows: using the BP neural network to extract effective defect feature values from each different type of defect image, constructing a corresponding input vector according to the structural characteristics of the BP neural network input layer; sending the input vector to the trained BP neural network, and the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron.

The display module is configured to: display the classification results to the user in the form of a graphical interface, and at the same time, store the classification results in a database for subsequent analysis and query.

A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the photovoltaic panel defect classification method based on color distribution and neural network as described in the first aspect of the present invention.

The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the photovoltaic panel defect classification method based on color distribution and neural network as described in the first aspect of the present invention are implemented.

One or more of the above technical solutions have the following beneficial effects:

(1) Before classifying the defects of photovoltaic panels, the present invention first converts the color space of the collected defect images of different types on the photovoltaic panels and converts the defect images from the RGB color space to the HSV color space through a color mapping method; quantizes the HSV color space and constructs a color distribution statistical graph, which is used to characterize the color distribution characteristics in the image. The defect images are pre-classified based on the color distribution characteristics in the defect images of photovoltaic panels, making full use of the fact that the color distribution characteristics of images of different defect types are also different, and can better distinguish the defect categories to which the photovoltaic panels belong.

(2) The present invention uses four different defect feature values as the input layer of the BP neural network. The combination of the four defect feature values can comprehensively reflect the defect characteristics of the defect image. At the same time, the present invention uses three different defect types as the output layer of the BP neural network. The input layer and the output layer establish a connection with the two hidden layers in the middle of the BP neural network through weights, which can not only realize the rapid identification of the defect category of the photovoltaic panel, but also ensure the accuracy of classification.

Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

FIG1 is a flow chart of a photovoltaic panel defect classification method based on color distribution and neural network in Embodiment 1 of the present invention.

FIG. 2 is a module schematic diagram of a photovoltaic panel defect classification system based on color distribution and neural network in Embodiment 2 of the present invention.

DETAILED DESCRIPTION

It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

Embodiment 1

As shown in FIG1 , this embodiment discloses a photovoltaic panel defect classification method based on color distribution and neural network, including:

Step S1, collecting images of different types of defects on photovoltaic panels.

Step S2, converting the acquired defect image into a color space, converting the defect image from the RGB color space into a representation of the HSV color space through a color mapping method; quantizing the HSV color space, and constructing a color distribution statistical graph, and using the color distribution statistical graph to characterize the characteristics of the color distribution in the image.

Step S3: pre-classify the defect image based on the color distribution characteristics, and the number of classifications is consistent with the defect types in the collected defect image.

Step S4, using the BP neural network to extract effective defect feature values from each different type of defect image, and construct a corresponding input vector according to the structural characteristics of the BP neural network input layer; the input vector is sent to the trained BP neural network, and the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron.

Step S5: Display the classification results to the user in the form of a graphical interface, and store the classification results in a database for subsequent analysis and query.

As an embodiment, a photovoltaic panel defect classification method based on color distribution and neural network is specifically as follows:

Step S1, collecting images of different types of defects on photovoltaic panels.

This embodiment collects images of different types of defects on photovoltaic panels. In order to better distinguish the types of defects of photovoltaic panels, defect images are collected based on three common and representative types of defects of photovoltaic panels, namely cracks, black blocks and hot spots.

In order to ensure the training effect while controlling the number of training samples during the subsequent BP neural network training, this embodiment collects 270 groups of each of the three different types of photovoltaic panel defect images, that is, a total of 810 groups of defect images as training samples for the subsequent BP neural network.

Step S2, converting the acquired defect image into a color space, converting the defect image from the RGB color space into a representation of the HSV color space through a color mapping method; quantizing the HSV color space, and constructing a color distribution statistical graph, and using the color distribution statistical graph to characterize the characteristics of the color distribution in the image.

Specifically, the method for converting the defect image from the RGB color space to the HSV color space by the color mapping method is:

;

in, and Respectively represent the maximum and minimum values of the color components in the three channels of R channel, G channel and B channel in the RGB color space; , and Respectively represent brightness, saturation, and hue in the HSV color space.

In the process of color space conversion in this embodiment, for the brightness in the HSV color space Instead of directly taking the maximum value of the RGB color components, the influence of different RGB color components on brightness is comprehensively considered.

Specifically, when quantizing the HSV color space, the brightness in the HSV color space is , Saturation And color tone They are quantized into 4 values, 4 values, and 16 values respectively; the quantized color space is expressed as:

;

in, represents the quantized color space, Represents the i- th quantized color in the quantized color space.

Specifically, the constructed color distribution statistical graph is expressed as:

;

in, Represents a color distribution statistical graph, Represents the ratio of the number of pixels of the i- th quantized color in the color distribution statistics graph, ; and Represents the width and height of the image respectively. represents the color quantization operation, represents a color mapping operation, represents the original defect image, It represents the i- th quantized color in the quantized color space.

Step S3: pre-classify the defect image based on the color distribution characteristics, and the number of classifications is consistent with the defect types in the collected defect image.

Specifically, according to the proportion of the number of pixels of various quantized colors in the color distribution statistics diagram, they are divided into three categories from low to high, corresponding to three different types of defects, namely: the first 270 groups of defect images with the lowest proportion of the number of pixels of quantized colors are divided into crack defect images, the first 270 groups of defect images with the highest proportion of the number of pixels of quantized colors are divided into hot spot defect images, and the remaining 270 groups of defect images are divided into black block defect images.

Step S4, using the BP neural network to extract effective defect feature values from each different type of defect image, and construct a corresponding input vector according to the structural characteristics of the BP neural network input layer; the input vector is sent to the trained BP neural network, and the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron.

Specifically, the BP neural network includes an input layer, two hidden layers and an output layer; the number of nodes in the input layer is consistent with the number of defect feature values, and the number of nodes in the output layer is consistent with the number of defect types in the defect image. The defect feature values include the number of defects, the area of the defect region, the grayscale mean of the defect region, and the aspect ratio of the defect region.

More specifically, the number of nodes in the input layer is 4, and the 4 nodes in the input layer correspond to the 4 defect feature values of defect number, defect area, defect area grayscale mean, and defect area aspect ratio. The two hidden layers in the BP neural network consist of the first hidden layer and the second hidden layer, which contain a total of 24 nodes, of which the first hidden layer and the second hidden layer contain 8 nodes and 16 nodes respectively. The number of nodes in the output layer of the BP neural network is 3, and the 3 nodes in the output layer correspond to the 3 defect types of cracks, black blocks, and hot spots.

The extracted defect feature values are constructed according to the structural characteristics of the BP neural network input layer to construct the corresponding input vector ;in, The corresponding number of defects, Corresponding to the defect area, Corresponding to the grayscale mean of the defect area, Corresponding to the aspect ratio of the defect area. After the input vector is sent to the trained BP neural network, the input vector passes through the weights and activation functions in the input layer and propagates forward layer by layer until it reaches the output layer.

Specifically, the input vector is connected to the first hidden layer through the first weight in the input layer, and to the second hidden layer through the second weight. After the output layer has been associated with the hidden layer, the third weight is used to establish a connection with the output layer and mark them with different label numbers, that is, crack defects correspond to label 1, black block defects correspond to label 2, and the hot spot defects correspond to label 3. The first weight is expressed as: , the second weight is expressed as: , the same reason, the third weight is expressed as: .

Finally, the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron. After the classification result is output, the classified image can be post-processed, such as smoothing, noise reduction, etc., to improve the stability of the classification result.

Step S5: Display the classification results to the user in the form of a graphical interface, and store the classification results in a database for subsequent analysis and query.

Before presenting the defects to the user in the form of a graphical interface, the location and category of the defects can be marked with boxes of different colors or other different types of symbols, and relevant statistical information can be provided to enhance the user experience.

Embodiment 2

As shown in FIG2 , this embodiment discloses a photovoltaic panel defect classification system based on color distribution and neural network, including:

The image acquisition module is configured to: acquire images of different types of defects on the photovoltaic panel.

The color distribution pre-classification module is configured to: perform color space conversion on the acquired defect image, convert the defect image from the RGB color space into the representation of the HSV color space through a color mapping method; quantify the HSV color space and construct a color distribution statistical graph, and use the color distribution statistical graph to characterize the characteristics of color distribution in the image; pre-classify the defect image based on the color distribution characteristics, and the number of classifications is consistent with the type of defects in the acquired defect image.

The defect image classification module is configured as follows: using the BP neural network to extract effective defect feature values from each different type of defect image, constructing a corresponding input vector according to the structural characteristics of the BP neural network input layer; sending the input vector to the trained BP neural network, and the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron.

The display module is configured to: display the classification results to the user in the form of a graphical interface, and at the same time, store the classification results in a database for subsequent analysis and query.

Embodiment 3

The purpose of this embodiment is to provide a computer-readable storage medium.

A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the photovoltaic panel defect classification method based on color distribution and neural network as described in the first embodiment of the present disclosure.

Embodiment 4

The purpose of this embodiment is to provide an electronic device.

An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the photovoltaic panel defect classification method based on color distribution and neural network as described in the first embodiment of the present disclosure are implemented.

The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Technical personnel in the relevant field should understand that various modifications or variations that can be made by technical personnel in the field without creative work on the basis of the technical solution of the present invention are still within the scope of protection of the present invention.

Claims (8)

1. A photovoltaic panel defect classification method based on color distribution and neural network, characterized by comprising: Collect images of different types of defects on photovoltaic panels; The collected defect image is converted into a color space, and the defect image is converted from the RGB color space into the HSV color space through a color mapping method; the HSV color space is quantified, and a color distribution statistical graph is constructed, and the color distribution statistical graph is used to characterize the characteristics of color distribution in the image; Specifically, when quantizing the HSV color space, the brightness in the HSV color space is , Saturation And color tone quantized into 4 values, 4 values, and 16 values respectively; The color distribution statistical diagram is expressed as: ; in, Represents a color distribution statistical graph, Represents the ratio of the number of pixels of the i- th quantized color in the color distribution statistics graph, ; and Represents the width and height of the image respectively. represents the color quantization operation, represents a color mapping operation, represents the original defect image, Then it represents the i- th quantized color in the quantized color space; The defect image is pre-classified based on the color distribution characteristics, and the number of classifications is consistent with the defect types in the collected defect image; Utilize BP neural network to extract effective defect feature values from each type of defect image, and construct corresponding input vectors according to the structural features of the BP neural network input layer; send the input vectors to the trained BP neural network, and the BP neural network obtains the classification result of the defect type to which the defect image belongs based on the output value of the neuron; The classification results are displayed to the user in the form of a graphical interface and stored in a database for subsequent analysis and query. 2. The photovoltaic panel defect classification method based on color distribution and neural network as described in claim 1 is characterized in that the defect types of the collected defect images on the photovoltaic panels include cracks, black blocks and hot spots. 3. The photovoltaic panel defect classification method based on color distribution and neural network according to claim 1, characterized in that the method for converting the defect image from the RGB color space to the HSV color space by the color mapping method is: ; in, and Respectively represent the maximum and minimum values of the color components in the three channels of R channel, G channel and B channel in the RGB color space; , and Respectively represent brightness, saturation, and hue in the HSV color space. 4. The photovoltaic panel defect classification method based on color distribution and neural network as described in claim 1 is characterized in that a BP neural network is used to extract effective defect feature values from each different type of defect image, and the defect feature values include the number of defects, the area of the defect area, the grayscale mean of the defect area, and the aspect ratio of the defect area. 5. The photovoltaic panel defect classification method based on color distribution and neural network as described in claim 1 is characterized in that the BP neural network comprises an input layer, two hidden layers and an output layer; the number of nodes in the input layer is consistent with the number of defect feature values, and the number of nodes in the output layer is consistent with the number of defect types in the defect image. 6. A photovoltaic panel defect classification system based on color distribution and neural network, characterized by comprising: The image acquisition module is configured to: acquire images of different types of defects on the photovoltaic panel; The color distribution pre-classification module is configured to: perform color space conversion on the acquired defect image, convert the defect image from the RGB color space into the representation of the HSV color space through a color mapping method; quantize the HSV color space and construct a color distribution statistical graph, and use the color distribution statistical graph to characterize the characteristics of color distribution in the image; pre-classify the defect image based on the color distribution characteristics, and the number of classifications is consistent with the defect types in the acquired defect image; specifically, when quantizing the HSV color space, the brightness in the HSV color space is converted into , Saturation And color tone quantized into 4 values, 4 values, and 16 values respectively; The color distribution statistical diagram is expressed as: ; in, Represents a color distribution statistical graph, Represents the ratio of the number of pixels of the i- th quantized color in the color distribution statistics graph, ; and Represents the width and height of the image respectively. represents the color quantization operation, represents a color mapping operation, represents the original defect image, Then it represents the i- th quantized color in the quantized color space; The defect image classification module is configured to: extract effective defect feature values from each type of defect image using a BP neural network, construct a corresponding input vector according to the structural features of the BP neural network input layer; send the input vector to the trained BP neural network, and the BP neural network obtains a classification result of the defect type to which the defect image belongs based on the output value of the neuron; The display module is configured to: display the classification results to the user in the form of a graphical interface, and at the same time, store the classification results in a database for subsequent analysis and query. 7. A computer-readable storage medium having a program stored thereon, wherein when the program is executed by a processor, the steps in the photovoltaic panel defect classification method based on color distribution and neural network as described in any one of claims 1 to 5 are implemented. 8. An electronic device, comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the photovoltaic panel defect classification method based on color distribution and neural network as described in any one of claims 1 to 5 are implemented.