Photovoltaic park vegetation phenology intelligent monitoring method and device based on digital camera

Abstract

The application discloses a digital camera-based photovoltaic park vegetation phenology intelligent monitoring method and device, relates to the technical field of ecological environment monitoring, and comprises the following steps: based on the camera point position of optimized arrangement, high-frequency collection of photovoltaic park vegetation growth images, semantic segmentation of the images, and obtaining segmentation masks of different micro-environment regions; through a radiation transmission model, the pixel values of different micro-environment regions are normalized and corrected, and an illumination correction image is generated; the corresponding multi-channel pixel values of each region are extracted at an edge end, the green chromaticity index is calculated, and the green chromaticity index is subjected to smoothing and noise reduction processing, so as to extract the key phenological periods of the vegetation in each region, and then the phenological difference index corresponding to the target photovoltaic park is calculated; the calculation result is uploaded to the cloud end; and finally, the ecological effect of photovoltaic facilities in the target photovoltaic park is quantified by the phenological difference index, thereby reducing the communication cost and realizing fine intelligent monitoring and quantitative evaluation of the vegetation phenology of different regions in the photovoltaic park.

Description

Technical Field

[0001] This application relates to the field of ecological environment monitoring technology, and in particular to a method and device for intelligent monitoring of vegetation phenology in photovoltaic parks based on digital cameras. Background Technology

[0002] Currently, the monitoring and assessment methods for the impact of photovoltaic development on vegetation phenology mainly rely on manual field surveys or satellite remote sensing technology. However, traditional manual surveys are mostly focused on reproductive phenology such as flowering, which is inefficient, lacks spatiotemporal continuity, and has a lack of systematic observation of leaf phenology. Satellite remote sensing technology has limited spatial resolution, cannot penetrate photovoltaic panels to observe vegetation in the shaded areas below them, and is difficult to effectively distinguish sub-regions with significant heterogeneity in the microenvironment within the photovoltaic array, making it difficult to conduct refined impact assessments at the module scale.

[0003] In recent years, methods based on near-ground remote sensing technologies (such as phenological digital cameras) have gradually become a research hotspot in this field, serving as a bridge between manual ground observation and satellite remote sensing. Compared with traditional methods, phenological digital cameras offer advantages such as high spatiotemporal resolution, low cost, and the ability to achieve continuous, non-destructive observation. By continuously capturing images and calculating greenness indices such as GCC (Green Chromatic Coordinate), continuous monitoring of phenology can be achieved. Currently, related technologies have made cutting-edge explorations in refining and expanding the spatiotemporal scope of phenological extraction using digital cameras. For example, existing technologies can identify and remove non-leaf background pixels such as soil and branches in digital camera images by calculating the amplitude of vegetation indexes, effectively improving the signal-to-noise ratio of the mean greenness index of the target vegetation, thereby enhancing the extraction accuracy of key phenological periods. Furthermore, existing technologies can also spatially match and regress GCC data extracted by ground-based phenological cameras with satellite remote sensing images, achieving high spatiotemporal resolution reconstruction of continuous vegetation index time series. These research findings have laid a solid methodological foundation for conducting high-frequency, long-term, and fine-scale vegetation phenology monitoring.

[0004] However, when applying existing digital camera technology directly to the specific scenario of photovoltaic parks, the following drawbacks still exist: 1. Specific lighting interference leads to severe data distortion: The shading of photovoltaic panels causes extremely uneven light distribution within the park, creating strong dynamic shadows; simultaneously, the glass surface of the photovoltaic modules strongly reflects sunlight. This results in significant physical interference between the RGB (Red, Green, Blue) values ​​of pixels in different areas (under the panels, between the panels, and the perimeter) of images captured at the same time, rather than simply reflecting the physiological state of the vegetation itself. The directly calculated GCC sequence is noisy and severely distorted. Although existing technologies provide methods for removing soil background pixels, they do not address the unique dynamic shadows and reflected light interference issues specific to photovoltaic systems.

[0005] 2. High hardware deployment costs and coverage blind spots: Photovoltaic parks have highly heterogeneous microenvironments. If traditional methods are used to fully cover these microenvironmental areas, a large number of dense camera arrays need to be deployed, resulting in high equipment procurement and maintenance costs.

[0006] 3. High Cost of Massive Image Transmission and Processing (Lack of Edge Computing): Continuous monitoring using high-frequency digital camera arrays generates massive amounts of high-definition image data daily. Transmitting all these raw images back to the cloud for processing would result in extremely high bandwidth pressure and communication costs (especially for photovoltaic power stations in remote deserts or high-altitude areas like the Qinghai-Tibet Plateau). There is also a high risk of network latency, and existing integrated monitoring systems lack a collaborative processing mechanism between "intelligent identification" and "edge computing."

[0007] In summary, in the existing technology, the interference of light and shadow in the complex environment of photovoltaic parks leads to low accuracy of phenological extraction, and the communication cost of transmitting massive amounts of high-definition images is high, which urgently needs to be solved. Summary of the Invention

[0008] This application provides a method and device for intelligent monitoring of vegetation phenology in photovoltaic parks based on digital cameras, in order to solve the problems in the prior art, such as low phenological extraction accuracy caused by light and shadow interference in complex photovoltaic environments, and high communication costs for transmitting massive amounts of high-definition images.

[0009] The first aspect of this application provides a method for intelligent monitoring of vegetation phenology in a photovoltaic park based on a digital camera, comprising the following steps: acquiring vegetation growth images of a target photovoltaic park, and performing a preset semantic segmentation operation on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images; based on the segmentation mask and a pre-constructed radiative transfer model, performing a preset normalization correction operation on the pixel values ​​of different microenvironment regions in the vegetation growth images to generate corresponding illumination correction images; extracting multi-channel pixel values ​​corresponding to each vegetation region in the illumination correction images, calculating the corresponding green chromaticity index sequence based on the multi-channel pixel values, and performing smoothing and noise reduction processing on the green chromaticity index sequence to extract the key phenological periods of each vegetation in the different microenvironment regions, and calculating the phenological difference index corresponding to the target photovoltaic park based on the key phenological periods, so as to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index.

[0010] Optionally, in one embodiment of this application, the step of acquiring vegetation growth images of the target photovoltaic park and performing a preset semantic segmentation operation on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images includes: determining the deployment optimization target corresponding to the target digital camera based on the image acquisition requirements corresponding to the target photovoltaic park; acquiring the array geometric parameters of the target photovoltaic park and inputting the array geometric parameters into a pre-constructed solar radiation transmission model, and combining the deployment optimization target to perform spatial simulation calculations to determine the optimal three-dimensional deployment location of the target digital camera, so that the target digital camera meets the preset area coverage requirements at the optimal three-dimensional deployment location; and at the optimal three-dimensional deployment location... The target digital camera is deployed at a designated location, and vegetation growth images of the target photovoltaic park are acquired through the target digital camera. The target digital camera meets the preset weather resistance requirements of its internal components. At a preset edge computing node, based on a pre-constructed semantic segmentation model, a preset semantic segmentation operation is performed on the vegetation growth images to obtain multiple category labels corresponding to the vegetation growth images and classification masks corresponding to the multiple category labels. Based on the multiple category labels and the classification masks, the corresponding segmentation mask is determined. The multiple category labels include photovoltaic panel body category, under-panel shaded vegetation area category, inter-panel vegetation area category, peripheral full-sunlight vegetation area category, bare soil category, and sky category.

[0011] Optionally, in one embodiment of this application, the step of performing a preset normalization correction operation on the pixel values ​​of different microenvironment regions in the vegetation growth image based on the segmentation mask and a pre-built radiative transfer model to generate a corresponding illumination correction image includes: at the edge computing node, determining the corresponding empirical model coefficients according to the segmentation mask, and obtaining the theoretical total irradiance of the pixels corresponding to the different microenvironment regions generated by the radiative transfer model; based on the theoretical total irradiance and the empirical model coefficients, performing a preset compensation operation on the original pixel brightness values ​​corresponding to the different microenvironment regions to obtain target pixel values ​​that meet preset vegetation color reflection requirements, so as to determine the illumination correction image based on the target pixel values.

[0012] Optionally, in one embodiment of this application, the step of extracting multi-channel pixel values ​​corresponding to each vegetation region in the illumination-corrected image, calculating the corresponding green chromaticity index sequence based on the multi-channel pixel values, and performing smoothing and noise reduction processing on the green chromaticity index sequence to extract the key phenological periods of each vegetation in the different microenvironment regions, and calculating the phenological difference index corresponding to the target photovoltaic park based on the key phenological periods, includes: converting the image matrix corresponding to the illumination-corrected image into corresponding floating-point data in the edge computing node, extracting the corresponding multi-channel pixel values ​​based on the floating-point data, and calculating the green chromaticity index corresponding to each vegetation region based on the multi-channel pixel values ​​and a preset vegetation region mask, and then using the... The green chromaticity index determines a green chromaticity index sequence corresponding to multiple vegetation areas; the green chromaticity index sequence is sent to a preset cloud server, where the cloud server performs smoothing and noise reduction processing on the green chromaticity index sequence to obtain a corresponding noise-reduced sequence, and a curve fitting operation is performed on the noise-reduced sequence to generate a corresponding fitted sequence. Based on the fitted sequence, the key phenological periods of each vegetation in each vegetation area are extracted, wherein the key phenological periods include the vegetation greening period and the vegetation yellowing period; the phenological period difference between different locations of the photovoltaic panel is calculated based on the key phenological periods, and the phenological difference index is calculated based on the phenological period difference, wherein the different locations of the photovoltaic panel include the area below the photovoltaic panel, the area between photovoltaic panels, and the area surrounding the photovoltaic panel.

[0013] Optionally, in one embodiment of this application, the mathematical expression for the target pixel value is:

[0014] in, The original pixel brightness values ​​of the pixels corresponding to the different microenvironment regions; This represents the theoretical total irradiance of the pixels corresponding to the different microenvironment regions. These are the empirical model coefficients corresponding to the various category labels; This represents the target pixel value.

[0015] Optionally, in one embodiment of this application, the mathematical expression for the phenological difference index is:

[0016] in, This indicates the accumulated days of the key phenological period corresponding to the area under the photovoltaic panel; This indicates the accumulated days of the key phenological period corresponding to the photovoltaic peripheral area; This indicates the accumulated days of the key phenological period corresponding to the inter-panel area; This represents the phenological difference index.

[0017] A second aspect of this application provides an intelligent monitoring device for vegetation phenology in a photovoltaic park based on a digital camera, comprising: a semantic segmentation module for acquiring vegetation growth images of a target photovoltaic park and performing a preset semantic segmentation operation on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images; a correction module for performing a preset normalization correction operation on the pixel values ​​of different microenvironment regions in the vegetation growth images based on the segmentation mask and a pre-constructed radiative transfer model to generate corresponding illumination correction images; and a quantification and evaluation module for extracting multi-channel pixel values ​​corresponding to each vegetation region in the illumination correction images, calculating the corresponding green chromaticity index sequence based on the multi-channel pixel values, and performing smoothing and noise reduction processing on the green chromaticity index sequence to extract the key phenological periods of each vegetation in the different microenvironment regions, and calculating the phenological difference index corresponding to the target photovoltaic park based on the key phenological periods, so as to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index.

[0018] Optionally, in one embodiment of this application, the semantic segmentation module includes: a first determining unit, used for determining the image acquisition requirements corresponding to the target photovoltaic park and determining the deployment optimization target corresponding to the target digital camera; a first acquiring unit, used for acquiring the array geometric parameters of the target photovoltaic park, inputting the array geometric parameters into a pre-constructed solar radiation transmission model, and combining the deployment optimization target to perform spatial simulation calculations to determine the optimal three-dimensional deployment site of the target digital camera, so that the target digital camera meets the preset area coverage requirements at the optimal three-dimensional deployment site; and an acquiring unit, used for deploying the target digital camera at the optimal three-dimensional deployment site and, through the... The target digital camera acquires vegetation growth images of the target photovoltaic park, wherein the target digital camera meets preset weather resistance requirements for its internal components; the second determining unit is used to perform preset semantic segmentation operations on the vegetation growth images at preset edge computing nodes based on a pre-constructed semantic segmentation model, to obtain multiple category labels corresponding to the vegetation growth images and classification masks corresponding to the multiple category labels, and to determine the corresponding segmentation mask based on the multiple category labels and the classification mask, wherein the multiple category labels include photovoltaic panel body category, under-panel shaded vegetation area category, inter-panel vegetation area category, peripheral full-sunlight vegetation area category, bare soil category, and sky category.

[0019] Optionally, in one embodiment of this application, the correction module includes: a second acquisition unit, configured to determine the corresponding empirical model coefficients at the edge computing node based on the segmentation mask, and acquire the theoretical total irradiance of the pixels corresponding to the different microenvironment regions generated by the radiative transfer model; and a compensation unit, configured to perform a preset compensation operation on the original pixel brightness values ​​corresponding to the different microenvironment regions based on the theoretical total irradiance and the empirical model coefficients, to obtain target pixel values ​​that meet preset vegetation color reflection requirements, so as to determine the illumination correction image based on the target pixel values.

[0020] Optionally, in one embodiment of this application, the quantization evaluation module includes: a conversion unit, configured to convert the image matrix corresponding to the illumination correction image into corresponding floating-point data in the edge computing node, extract corresponding multi-channel pixel values ​​based on the floating-point data, calculate the green chromaticity index corresponding to each vegetation region based on the multi-channel pixel values ​​and a preset vegetation region mask, and determine a green chromaticity index sequence corresponding to multiple vegetation regions through the green chromaticity index; and a noise reduction unit, configured to send the green chromaticity index sequence to a preset cloud server for quantization. The end server performs smoothing and noise reduction processing on the green chromaticity index sequence to obtain the corresponding noise-reduced sequence, and performs curve fitting operation on the noise-reduced sequence to generate the corresponding fitted sequence. Based on the fitted sequence, the key phenological periods of each vegetation in each vegetation area are extracted, wherein the key phenological periods include the vegetation greening period and the vegetation yellowing period. The calculation unit is used to calculate the phenological period difference between different locations of the photovoltaic panel according to the key phenological periods, and calculate the phenological difference index according to the phenological period difference, wherein the different locations of the photovoltaic panel include the area below the photovoltaic panel, the area between photovoltaic panels, and the area surrounding the photovoltaic panel.

[0021] Optionally, in one embodiment of this application, the mathematical expression for the target pixel value is:

[0022] in, The original pixel brightness values ​​of the pixels corresponding to the different microenvironment regions; This represents the theoretical total irradiance of the pixels corresponding to the different microenvironment regions. These are the empirical model coefficients corresponding to the various category labels; This represents the target pixel value.

[0023] Optionally, in one embodiment of this application, the mathematical expression for the phenological difference index is:

[0024] in, This indicates the accumulated days of the key phenological period corresponding to the area under the photovoltaic panel; This indicates the accumulated days of the key phenological period corresponding to the photovoltaic peripheral area; This indicates the accumulated days of the key phenological period corresponding to the inter-panel area; This represents the phenological difference index.

[0025] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras as described in the above embodiments.

[0026] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent monitoring method for vegetation phenology in a photovoltaic park based on a digital camera.

[0027] A fifth aspect of this application provides a computer program product, including a computer program that is executed to implement the above-described intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras.

[0028] Therefore, the embodiments of this application have the following beneficial effects: The embodiments of this application can acquire vegetation growth images of a target photovoltaic park and perform preset semantic segmentation operations on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images. Based on the segmentation mask and a pre-constructed radiative transfer model, preset normalization correction operations are performed on the pixel values ​​of different microenvironment regions in the vegetation growth images to generate corresponding illumination correction images. Multi-channel pixel values ​​corresponding to each vegetation region in the illumination correction images are extracted, and the corresponding green chromaticity index sequence is calculated based on the multi-channel pixel values. The green chromaticity index sequence is then smoothed and denoised to extract the key phenological periods of each vegetation in different microenvironment regions. Based on the key phenological periods, the phenological difference index corresponding to the target photovoltaic park is calculated to quantify the ecological effects of photovoltaic facilities in the target photovoltaic park through the phenological difference index. This application can improve the accuracy of phenological extraction; at the same time, by optimizing the deployment and introducing edge computing technology with "end-edge-cloud" collaboration, communication costs are reduced, realizing refined intelligent monitoring and quantitative evaluation of vegetation phenology in different areas of the photovoltaic park. This solves the problems in existing technologies, such as low accuracy of phenological extraction due to light and shadow interference in the complex environment of photovoltaic parks, and high communication costs for transmitting massive amounts of high-definition images.

[0029] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0030] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1This is a flowchart illustrating an intelligent monitoring method for vegetation phenology in a photovoltaic park based on a digital camera, according to an embodiment of this application. Figure 2 This is an example diagram of a digital camera-based intelligent monitoring device for vegetation phenology in a photovoltaic park, according to an embodiment of this application. Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0031] Among them, 10-Intelligent monitoring device for vegetation phenology in photovoltaic parks based on digital cameras, 100-Semantic segmentation module, 200-Correction module, 300-Quantitative evaluation module, 301-Memory, 302-Processor, and 303-Communication interface. Detailed Implementation

[0032] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0033] The following describes, with reference to the accompanying drawings, an intelligent monitoring method and apparatus for vegetation phenology in photovoltaic parks based on digital cameras, according to embodiments of this application. Addressing the problems mentioned in the background art, this application provides an intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras. In this method, images of plant growth in the photovoltaic park are acquired at high frequency based on optimized camera placement, and semantically segmented to obtain segmentation masks for different regions. A radiative transfer model is used to normalize and correct pixel values ​​in different microenvironment regions, generating illumination-corrected images. Multi-channel pixel values ​​corresponding to each vegetation region are extracted at the edge, and a green chromaticity index is calculated and smoothed and denoised to extract the key phenological periods of vegetation in each region. Then, the phenological difference index corresponding to the target photovoltaic park is calculated, and the calculation results are uploaded to the cloud. Finally, the ecological effects of photovoltaic facilities in the target photovoltaic park are quantified using the phenological difference index, thereby reducing communication costs and achieving refined intelligent monitoring and quantitative assessment of vegetation phenology in different regions of the photovoltaic park. This solves the problems in the prior art, such as low phenological extraction accuracy due to light and shadow interference in the complex environment of photovoltaic parks, and high communication costs for transmitting massive amounts of high-definition images.

[0034] Specifically, Figure 1 This is a flowchart illustrating an intelligent monitoring method for vegetation phenology in a photovoltaic park based on a digital camera, provided as an embodiment of this application.

[0035] like Figure 1 As shown, the intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras includes the following steps: In step S101, vegetation growth images of the target photovoltaic park are acquired, and a preset semantic segmentation operation is performed on the vegetation growth images to obtain the segmentation mask corresponding to the vegetation growth images. The embodiments of this application firstly determine the optimal three-dimensional deployment location of the digital camera, and then deploy the digital camera at the optimal three-dimensional deployment location to intelligently acquire vegetation growth images of the photovoltaic park, i.e., a single RGB image; secondly, the embodiments of this application can perform intelligent recognition and region segmentation operations on the single RGB image to obtain the segmentation mask corresponding to the single RGB image.

[0036] Optionally, in one embodiment of this application, vegetation growth images of the target photovoltaic park are acquired, and a preset semantic segmentation operation is performed on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images. This includes: determining the deployment optimization target corresponding to the target digital camera based on the image acquisition requirements of the target photovoltaic park; obtaining the array geometric parameters of the target photovoltaic park, and inputting the array geometric parameters into a pre-constructed solar radiation transmission model, and combining the deployment optimization target to perform spatial simulation calculations to determine the optimal three-dimensional deployment site of the target digital camera, so that the target digital camera meets the preset area coverage requirements at the optimal three-dimensional deployment site; and performing spatial simulation calculations at the optimal three-dimensional deployment site. A target digital camera is deployed at a designated location to acquire vegetation growth images of the target photovoltaic park. The target digital camera meets the preset weather resistance requirements of its internal components. At a preset edge computing node, based on a pre-built semantic segmentation model, a preset semantic segmentation operation is performed on the vegetation growth images to obtain multiple category labels and classification masks corresponding to the vegetation growth images. Based on the multiple category labels and classification masks, the corresponding segmentation mask is determined. The multiple category labels include photovoltaic panel body category, under-panel shaded vegetation area category, inter-panel vegetation area category, peripheral full-sunlight vegetation area category, bare soil category, and sky category.

[0037] It should be noted that, in this embodiment, the array geometric parameters of the target photovoltaic park (including the tilt angle, azimuth angle, array spacing, installation height, etc. of the photovoltaic panels) can be obtained first, and then input into the solar radiation transfer model (such as the Perez model); secondly, this embodiment takes "maximizing the coverage of multiple types of heterogeneous areas by a single camera" as the optimization objective (i.e., the deployment optimization objective), performs spatial simulation calculations, and determines the optimal three-dimensional deployment position of the digital camera (such as latitude and longitude, installation height, pitch angle, and yaw angle, etc.), thereby ensuring that the field of view of a single camera simultaneously includes the shadow area under the photovoltaic panels and the vegetation area between the panels.

[0038] As one possible approach, to enable a single digital camera to simultaneously cover the shaded vegetation area under the slab, the vegetation area between the slabs, and the outer fully illuminated vegetation area within a limited field of view, the camera deployment parameters in this embodiment can be expressed as follows: ,in, Indicates the camera installation location. Indicates pitch angle, Indicates the yaw angle. This represents the focal length or equivalent field of view parameter, and the objective function for optimizing this layout is... The mathematical expression is:

[0039] in, Indicates the overall coverage rate of multiple microenvironment areas; This indicates the average effective pixel resolution of the target area; Indicates the occlusion rate; Indicates the risk value of direct sunlight glare; This represents the weighting coefficient.

[0040] To ensure that the under-plate region, the inter-plate region, and the outer region are all effectively observed, the embodiments of this application may further set coverage constraints. ,in, Indicates the first Class target region; Indicates the image imaging range; Indicates the minimum coverage threshold for the corresponding area; Indicates the first Class target region Camera deployment parameter set The set of pixels projected onto the image plane.

[0041] Furthermore, in this embodiment, a modified ordinary digital camera can be deployed at the optimal location. This camera is equipped with an independent solar power supply module and enhances the weather resistance of its internal components (i.e., meets the preset weather resistance requirements of internal components) to adapt to signal transmission and stable operation under extreme conditions such as the Qinghai-Tibet Plateau, arid deserts, or rain and snow. In addition, the camera has a built-in timed wake-up program that automatically performs continuous shooting at a preset frequency.

[0042] Subsequently, in the image processing of this embodiment, a lightweight deep learning semantic segmentation model (such as a pre-trained DeepLabV3+ model) can be introduced to perform localized automatic recognition and accurate segmentation on the acquired single RGB image, so as to automatically divide the image pixels into category labels such as "photovoltaic panel body", "under-panel shaded vegetation area", "inter-panel vegetation area", "peripheral full-sunlight vegetation area", "bare soil" and "sky", and generate corresponding classification masks to determine the corresponding segmentation masks, thereby providing a basis for differentiated processing.

[0043] In practical implementation, embodiments of this application can input a single RGB image into a semantic segmentation model to output a corresponding pixel-level category probability map, and determine the corresponding category label based on the probability value of each pixel in different categories. Specifically, embodiments of this application can represent the input image as follows: ,in, Indicates the image height. The image width is represented by 3, and the three color channels (red, green, and blue) are represented by 3. The class probability tensor output by the semantic segmentation model can be expressed as: ,in, Indicates the total number of preset categories; Represents pixels in an image Belongs to the The probability value of the class.

[0044] Furthermore, in this embodiment of the application, pixel-level category labels can be determined based on the category probability values ​​corresponding to each pixel. The specific expression can be represented as follows:

[0045] in, Represents pixels ( i , j Category labels.

[0046] Furthermore, based on the category labels of all pixels, this application embodiment can generate classification masks corresponding to the photovoltaic panel body area, the under-panel shaded vegetation area, the inter-panel vegetation area, the peripheral full-sunlight vegetation area, the bare soil area, and the sky area, and determine the corresponding segmentation mask.

[0047] It should be noted that, to improve the spatial continuity and boundary accuracy of the segmentation results, this embodiment of the application can also perform post-processing operations on the classification mask. These post-processing operations include, but are not limited to, connected component filtering, morphological opening operations, morphological closing operations, small hole filling, and boundary smoothing, in order to remove isolated noise pixels, eliminate local mis-segmented regions, and maintain the stability of the target region boundary. The post-processed classification mask can serve as the regional constraint basis for subsequent illumination correction and vegetation index extraction.

[0048] Furthermore, to avoid interference from the sky area, the photovoltaic panel body area, and the exposed soil area in vegetation index calculation, this embodiment retains only the vegetation category masks corresponding to the shaded vegetation area under the panel, the inter-panel vegetation area, and the outer full-sunlight vegetation area, and determines the effective vegetation pixel range for subsequent statistical purposes based on the vegetation category masks. Therefore, this embodiment can achieve automatic positioning, accurate segmentation, and stable extraction of vegetation areas in different microenvironments, providing a reliable foundation for subsequent differential correction of pixel values ​​in different areas, calculation of green chromaticity index, and extraction of key phenological periods.

[0049] For continuous images collected at adjacent time points from the same monitoring point, this application embodiment can also combine the spatial overlap relationship between regions of the same category at adjacent time points to perform consistency correction on the classification mask at the current time point, so as to reduce the impact of instantaneous illumination disturbance, local occlusion or noise fluctuation on the stability of segmentation results, thereby improving the consistency of time series vegetation region extraction.

[0050] It is understandable that the embodiments of this application break away from the traditional dense deployment pattern. By jointly optimizing the three-dimensional spatial geometric parameters and the solar radiation model, it achieves "single camera coverage of multiple heterogeneous microhabitat areas", achieving the optimal monitoring area with the fewest number of devices. The hardware investment cost is low and the cost performance is extremely high.

[0051] Therefore, the embodiments of this application innovatively introduce a collaborative architecture of intelligent recognition and edge computing, which pre-completes the heavy image processing (segmentation, correction, exponential calculation) locally, and transmits only a very small amount of feature data to the cloud, effectively improving data transmission efficiency, reducing operation and maintenance, and completely solving the problems of limited network bandwidth, high communication costs (i.e. hardware costs and adaptability to extreme environments) and difficulty in finely distinguishing microenvironments in arid / remote photovoltaic power stations.

[0052] In step S102, based on the segmentation mask and the pre-built radiative transfer model, the pixel values ​​of different microenvironment regions in the vegetation growth image are subjected to a preset normalization correction operation to generate the corresponding illumination correction image. Furthermore, in this embodiment, the pixel values ​​of different microenvironment regions (especially the under-panel shadow area and the area affected by reflected light) can be normalized and corrected based on the segmentation mask obtained above and combined with the theoretical irradiance data synchronously output by the radiative transfer model.

[0053] Therefore, the embodiments of this application effectively solve problems such as special lighting distortion through anti-interference lighting correction operations.

[0054] Optionally, in one embodiment of this application, based on a segmentation mask and a pre-built radiative transfer model, a preset normalization correction operation is performed on the pixel values ​​of different microenvironment regions in a vegetation growth image to generate a corresponding illumination correction image. This includes: at an edge computing node, determining the corresponding empirical model coefficients based on the segmentation mask, and obtaining the theoretical total irradiance of the pixels corresponding to different microenvironment regions generated by the radiative transfer model; based on the theoretical total irradiance and the empirical model coefficients, performing a preset compensation operation on the original pixel brightness values ​​corresponding to different microenvironment regions to obtain target pixel values ​​that meet preset vegetation color reflection requirements, so as to determine the illumination correction image based on the target pixel values.

[0055] Specifically, embodiments of this application can use theoretical total irradiance to compensate and correct the brightness values ​​of the disturbed original pixels, eliminate local overexposure or underexposure caused by dynamic shadows and glass reflections, and extract pixel values ​​that truly reflect the physiological color of the vegetation itself (i.e., target pixel values ​​that meet the preset vegetation color reflection requirements).

[0056] As one possible approach, the theoretical total irradiance in the embodiments of this application can be decomposed into the sum of direct irradiance, sky diffuse irradiance, and environmental reflected irradiance to enhance the characterization of the combined effects of dynamic shading of photovoltaic panels and glass reflection. As shown in the following formula:

[0057] in, Represents pixels At any moment The corresponding direct irradiance component; Represents the sky-scattered irradiance component; This represents the reflected irradiance component caused by the surface of the photovoltaic panel and the surrounding ground. Based on the pixel's category label. In this embodiment, differential correction can be performed on the red, green, and blue channels separately. The mathematical expression for the correction process is as follows:

[0058] in, , and These represent the channel gain coefficient and correction index related to the region category, respectively; Indicates reference irradiance; This indicates the minimum value to prevent the denominator from being zero.

[0059] Furthermore, embodiments of this application can also suppress or eliminate highly reflective pixels based on a high-brightness mask to prevent locally overexposed pixels caused by reflection from photovoltaic module glass from participating in vegetation index statistics.

[0060] Optionally, in one embodiment of this application, the mathematical expression for the target pixel value is:

[0061] in, The original pixel brightness values ​​for pixels corresponding to different microenvironment regions; The theoretical total irradiance of pixels corresponding to different microenvironment regions; These are the empirical model coefficients corresponding to multiple category labels; Indicates the target pixel value.

[0062] It should be noted that, in the embodiments of this application, the mathematical expression for the illumination correction process (i.e., the target pixel value) is:

[0063] in, This represents the original pixel brightness value. This is the theoretical total irradiance corresponding to that pixel (i.e., the theoretical total irradiance of the pixel corresponding to the microenvironment region). These are the empirical model coefficients calibrated based on the region segmentation categories; Indicates the target pixel value.

[0064] Therefore, the embodiments of this application have customized a physical model-driven illumination correction algorithm for photovoltaic parks (that is, combining the original pixel brightness value with the theoretical total irradiance of the radiative transfer model to compensate for the interference of photovoltaic panel shadows and glass reflections). This eliminates the interference of dynamic strong shadows and strong reflective light from photovoltaic panels at the pixel level, restores the true physiological color characteristics of vegetation (i.e., green color index), ensures the scientificity and reliability of phenological extraction, and significantly improves the accuracy of ecological monitoring.

[0065] In step S103, multi-channel pixel values ​​corresponding to each vegetation area in the illumination-corrected image are extracted, and the corresponding green chromaticity index sequence is calculated based on the multi-channel pixel values. The green chromaticity index sequence is then smoothed and denoised to extract the key phenological periods of each vegetation in different microenvironment areas. Based on the key phenological periods, the phenological difference index corresponding to the target photovoltaic park is calculated to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index. Subsequently, embodiments of this application can extract multi-channel pixel values ​​of each vegetation region in the illumination-corrected image, calculate the corresponding green chromaticity index, and perform smoothing and noise reduction processing on multiple green chromaticity index sequences corresponding to each vegetation region to extract the key phenological periods of each vegetation, calculate the phenological difference index corresponding to the photovoltaic park, and thus quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index.

[0066] Therefore, the embodiments of this application are based on edge computing image processing and data compression strategies. They use deep learning models to segment microhabitats at the edge and directly calculate GCC / EXG (Excess Green Index) feature values. The lightweight mechanism of transmitting only the calculation results and timestamps to the cloud instead of the original images overcomes the bottleneck of massive data transmission and solves the problem of quantitative evaluation gaps.

[0067] Optionally, in one embodiment of this application, multi-channel pixel values ​​corresponding to each vegetation region in the illumination-corrected image are extracted, and the corresponding green chromaticity index sequence is calculated based on the multi-channel pixel values. The green chromaticity index sequence is then smoothed and denoised to extract the key phenological periods of each vegetation in different microenvironment regions. The phenological difference index corresponding to the target photovoltaic park is calculated based on the key phenological periods. This includes: converting the image matrix corresponding to the illumination-corrected image into corresponding floating-point data in the edge computing node, extracting the corresponding multi-channel pixel values ​​based on the floating-point data, and calculating the green chromaticity index corresponding to each vegetation region based on the multi-channel pixel values ​​and a preset vegetation region mask. The chromaticity index determines the green chromaticity index sequence corresponding to multiple vegetation areas. The green chromaticity index sequence is sent to a preset cloud server for smoothing and noise reduction processing to obtain the corresponding noise-reduced sequence. Curve fitting is then performed on the noise-reduced sequence to generate the corresponding fitted sequence. Based on the fitted sequence, the key phenological periods of each vegetation in each vegetation area are extracted. The key phenological periods include the vegetation greening period and the vegetation yellowing period. The phenological period difference between different locations of the photovoltaic panel is calculated based on the key phenological periods, and the phenological difference index is calculated based on the phenological period difference. The different locations of the photovoltaic panel include the area under the photovoltaic panel, the area between photovoltaic panels, and the area surrounding the photovoltaic panel.

[0068] It should be noted that the above-mentioned image intelligent recognition and illumination correction tasks are deployed directly on edge computing nodes (i.e., cameras with computing power modules or local gateways) close to the data source.

[0069] After the edge correction is completed, the embodiments of this application can directly extract the red, green and blue three-channel pixel values ​​of each vegetation area and calculate the green color index.

[0070] Specifically, embodiments of this application can convert the image matrix into floating-point data and separate the R, G, and B bands, as shown in the following formula:

[0071] It should be noted that the embodiments of this application can target specific areas. At any moment Effective vegetation pixel set The average green chromaticity index of the region is calculated using the following formula:

[0072] in, , and These represent the red, green, and blue channel pixel values ​​after illumination correction, respectively. To ensure the reliability of the time sequence, embodiments of this application may further define the percentage of effective pixels in a region. ,in, Indicates the region The set of all vegetation pixels participating in the candidate statistics; when If the data is below a preset threshold, the data in the corresponding region at that time will be considered invalid samples.

[0073] Secondly, in this embodiment, the color index (i.e., green chromaticity index) of the target area (i.e., each vegetation area) can be calculated by combining a preset vegetation area mask, and pixel-level and area-level quality control can be carried out simultaneously to ensure the reliability of the green chromaticity index.

[0074] Furthermore, after calculating the effective GCC mean of each region (underboard, betweenboards, and periphery) at the edge nodes, this embodiment of the application only sends the lightweight structural data (calculated numerical results and corresponding timestamps) to the cloud server via the Internet of Things. In principle, the original high-definition images are not retained locally for a long time, and the lightweight result data is only retained after the index extraction is completed.

[0075] It should be noted that in the embodiments of this application, for scenarios such as abnormal data quality, abnormal device operation, or abnormal communication, a local temporary backup or abnormal image back transmission mechanism can be triggered to balance transmission efficiency and data traceability, thereby significantly reducing the communication burden caused by continuous back transmission of the original image, and compressing the uploaded content from high-definition images into lightweight structured data containing vegetation index values ​​and timestamps, which greatly reduces network bandwidth usage and cloud storage pressure.

[0076] Optionally, in one embodiment of this application, the mathematical expression for the phenological difference index is:

[0077] in, This indicates the accumulated days of the key phenological period corresponding to the area under the photovoltaic panel; This indicates the accumulated days of the key phenological period corresponding to the photovoltaic peripheral area; This indicates the accumulated days of the key phenological period corresponding to the inter-panel area; This represents the phenological difference index.

[0078] Those skilled in the art should understand that, in order to improve the stability of extraction during key phenological periods, the embodiments of this application can use a double Logistic model to fit the smoothed green chromaticity index sequence, as shown in the following formula:

[0079] The first term characterizes the spring greening and rising process, while the second term characterizes the autumn decline process. Indicates the region The corresponding fitting parameters. Based on the fitted curve, the greening-up period and the yellowing-out period can be extracted. For example, the greening-up period can be defined as... .

[0080] In some embodiments, to enhance the interpretability of differences in key phenological periods across different regions, the phenological difference index may also be in a weighted normalized form, as shown in the following equation:

[0081] in, Indicates the phenological stage to be evaluated; , and These represent the accumulated days of the year in the sub-plate area, inter-plate area, and outer area during the corresponding key phenological stages, respectively. This represents the weighting coefficient.

[0082] Subsequently, after receiving time-series GCC data (i.e., green color index sequence) from edge nodes on the cloud server, this embodiment of the application can apply the Savitzky-Golay filtering algorithm for smoothing and noise reduction based on the corrected vegetation index sequence, and use logistic curve fitting to accurately extract key phenological periods such as the start of season (SOS) and end of season (EOS) of vegetation.

[0083] Subsequently, embodiments of this application can calculate the phenological period differences under the photovoltaic panel, between photovoltaic panels, and in the photovoltaic periphery area, construct and calculate the Phenology Difference Index (PDI) to quantify the microclimate effect of photovoltaic facilities. The mathematical expression of the PDI (i.e., the Phenology Difference Index assessment model) is as follows:

[0084] in, , and These represent the Day of Year (DOY) of specific phenological periods (such as SOS) extracted from the sub-plate region, inter-plate region, and peripheral full-light region, respectively. Therefore, the embodiments of this application fill the gap in quantitative ecological assessment indicators by integrating the differences in phenological periods under the panel, between the panels, and in the surrounding full-light area to quantitatively evaluate the photovoltaic ecological effects. This not only realizes automated intelligent extraction of phenological periods but also proposes a phenological difference index, providing the scientific and industrial communities with a direct and quantitative mathematical tool for assessing the "microclimate region and ecological effects" generated by the construction of photovoltaic power plants.

[0085] In summary, the embodiments of this application, through a complete end-edge-cloud collaborative process including deployment optimization, intelligent acquisition of edge images, region recognition and segmentation, illumination correction, extraction of vegetation indices at the edge, and cloud-based assessment of phenological differences, can effectively achieve intelligent monitoring of vegetation phenology in photovoltaic parks and accurately quantify the ecological effects of photovoltaic facilities.

[0086] The intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras proposed in this application involves acquiring vegetation growth images of the target photovoltaic park and performing a preset semantic segmentation operation on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images. Based on the segmentation mask and a pre-constructed radiative transfer model, a preset normalization correction operation is performed on the pixel values ​​of different microenvironment regions in the vegetation growth images to generate corresponding illumination-corrected images. Multi-channel pixel values ​​corresponding to each vegetation region in the illumination-corrected images are extracted, and the corresponding green chromaticity index sequence is calculated based on the multi-channel pixel values. The green chromaticity index sequence is then smoothed and denoised to extract the key phenological periods of each vegetation in different microenvironment regions. Based on the key phenological periods, the phenological difference index corresponding to the target photovoltaic park is calculated to quantify the ecological effects of photovoltaic facilities in the target photovoltaic park through the phenological difference index. This application can improve the accuracy of phenological extraction; simultaneously, by optimizing the deployment and introducing edge computing technology with "end-edge-cloud" collaboration, communication costs are reduced, enabling refined intelligent monitoring and quantitative evaluation of vegetation phenology in different areas of the photovoltaic park.

[0087] Secondly, with reference to the accompanying drawings, an intelligent monitoring device for vegetation phenology in a photovoltaic park based on a digital camera, according to an embodiment of this application, is described.

[0088] Figure 2 This is a block diagram of a photovoltaic park vegetation phenology intelligent monitoring device based on a digital camera, according to an embodiment of this application.

[0089] like Figure 2 As shown, the intelligent monitoring device 10 for vegetation phenology in photovoltaic parks based on digital cameras includes: a semantic segmentation module 100, a correction module 200, and a quantitative evaluation module 300.

[0090] The semantic segmentation module 100 is used to acquire vegetation growth images of the target photovoltaic park and perform preset semantic segmentation operations on the vegetation growth images to obtain the segmentation mask corresponding to the vegetation growth images.

[0091] The correction module 200 is used to perform preset normalization correction operations on the pixel values ​​of different microenvironment regions in the vegetation growth image based on the segmentation mask and the pre-built radiative transfer model, so as to generate the corresponding illumination correction image.

[0092] The quantitative evaluation module 300 is used to extract the multi-channel pixel values ​​corresponding to each vegetation area in the illumination-corrected image, calculate the corresponding green chromaticity index sequence based on the multi-channel pixel values, and perform smoothing and noise reduction processing on the green chromaticity index sequence to extract the key phenological periods of each vegetation in different microenvironment areas. Based on the key phenological periods, the module calculates the phenological difference index corresponding to the target photovoltaic park, so as to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index.

[0093] Optionally, in one embodiment of this application, the semantic segmentation module 100 includes: a first determining unit, a first acquiring unit, a collection unit, and a second determining unit.

[0094] The first determining unit is used to determine the image acquisition requirements of the target photovoltaic park and to determine the deployment optimization target of the target digital camera.

[0095] The first acquisition unit is used to acquire the array geometric parameters of the target photovoltaic park, input the array geometric parameters into the pre-constructed solar radiation transmission model, and perform spatial simulation calculations in conjunction with the deployment optimization target to determine the optimal three-dimensional deployment position of the target digital camera, so that the target digital camera meets the preset area coverage requirements at the optimal three-dimensional deployment position.

[0096] The acquisition unit is used to deploy target digital cameras at the optimal three-dimensional deployment sites and acquire vegetation growth images of the target photovoltaic park through the target digital cameras. The target digital cameras meet the preset weather resistance requirements of the internal components.

[0097] The second determining unit is used to perform a preset semantic segmentation operation on the vegetation growth image at a preset edge computing node based on a pre-built semantic segmentation model, so as to obtain multiple category labels and classification masks corresponding to the vegetation growth image, and determine the corresponding segmentation mask based on the multiple category labels and classification masks. The multiple category labels include photovoltaic panel body category, under-panel shaded vegetation area category, inter-panel vegetation area category, peripheral full-sunlight vegetation area category, bare soil category, and sky category.

[0098] Optionally, in one embodiment of this application, the correction module 200 includes a second acquisition unit and a compensation unit.

[0099] The second acquisition unit is used to determine the corresponding empirical model coefficients based on the segmentation mask at the edge computing node, and to acquire the theoretical total irradiance of the pixels corresponding to different micro-environment regions generated by the radiative transfer model.

[0100] The compensation unit is used to perform preset compensation operations on the original pixel brightness values ​​corresponding to different microenvironment areas based on the theoretical total irradiance and empirical model coefficients, so as to obtain the target pixel value that meets the preset vegetation color reflection requirements, and to determine the illumination correction image based on the target pixel value.

[0101] Optionally, in one embodiment of this application, the quantization evaluation module 300 includes: a conversion unit, a noise reduction unit, and a calculation unit.

[0102] The conversion unit is used to convert the image matrix corresponding to the illumination correction image into the corresponding floating-point data in the edge computing node, extract the corresponding multi-channel pixel values ​​based on the floating-point data, and calculate the green chromaticity index corresponding to each vegetation area based on the multi-channel pixel values ​​and the preset vegetation area mask, and determine the green chromaticity index sequence corresponding to multiple vegetation areas through the green chromaticity index.

[0103] The noise reduction unit is used to send the green chromaticity index sequence to a preset cloud server, so that the cloud server can perform smooth noise reduction on the green chromaticity index sequence to obtain the corresponding noise reduction sequence. Then, the noise reduction sequence is subjected to curve fitting to generate the corresponding fitting sequence. Based on the fitting sequence, the key phenological periods of each vegetation in each vegetation area are extracted. The key phenological periods include the vegetation greening period and the vegetation withering period.

[0104] The calculation unit is used to calculate the phenological period difference between different locations of the photovoltaic panel based on the key phenological period, and to calculate the phenological difference index based on the phenological period difference. The different locations of the photovoltaic panel include the area below the photovoltaic panel, the area between photovoltaic panels, and the area surrounding the photovoltaic panel.

[0105] Optionally, in one embodiment of this application, the mathematical expression for the target pixel value is:

[0106] in, The original pixel brightness values ​​for pixels corresponding to different microenvironment regions; The theoretical total irradiance of pixels corresponding to different microenvironment regions; These are the empirical model coefficients corresponding to multiple category labels; Indicates the target pixel value.

[0107] Optionally, in one embodiment of this application, the mathematical expression for the phenological difference index is:

[0108] in, This indicates the accumulated days of the key phenological period corresponding to the area under the photovoltaic panel; This indicates the accumulated days of the key phenological period corresponding to the photovoltaic peripheral area; This indicates the accumulated days of the key phenological period corresponding to the inter-panel area; This represents the phenological difference index.

[0109] It should be noted that the foregoing explanation of the embodiment of the intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras also applies to the intelligent monitoring device for vegetation phenology in photovoltaic parks based on digital cameras in this embodiment, and will not be repeated here.

[0110] The intelligent monitoring device for vegetation phenology in photovoltaic parks based on digital cameras proposed in this application includes a semantic segmentation module 100, used to acquire vegetation growth images of the target photovoltaic park and perform preset semantic segmentation operations on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images; a correction module 200, used to perform preset normalization correction operations on the pixel values ​​of different microenvironment areas in the vegetation growth images based on the segmentation mask and a pre-constructed radiative transfer model to generate corresponding illumination correction images; and a quantification and evaluation module 300, used to extract the multi-channel pixel values ​​corresponding to each vegetation area in the illumination correction image, calculate the corresponding green chromaticity index sequence based on the multi-channel pixel values, and perform smoothing and noise reduction processing on the green chromaticity index sequence to extract the key phenological periods of each vegetation in different microenvironment areas, and calculate the phenological difference index corresponding to the target photovoltaic park based on the key phenological periods, so as to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index. This application can improve the accuracy of phenological extraction; at the same time, by optimizing the deployment and introducing edge computing technology that integrates "end-edge-cloud", it can reduce communication costs and realize refined intelligent monitoring and quantitative assessment of vegetation phenology in different areas of the photovoltaic park.

[0111] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 301, the processor 302, and the computer program stored on the memory 301 and capable of running on the processor 302.

[0112] When the processor 302 executes the program, it implements the intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras provided in the above embodiments.

[0113] Furthermore, electronic devices also include: Communication interface 303 is used for communication between memory 301 and processor 302.

[0114] The memory 301 is used to store computer programs that can run on the processor 302.

[0115] The memory 301 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0116] If the memory 301, processor 302, and communication interface 303 are implemented independently, then the communication interface 303, memory 301, and processor 302 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0117] Optionally, in a specific implementation, if the memory 301, processor 302, and communication interface 303 are integrated on a single chip, then the memory 301, processor 302, and communication interface 303 can communicate with each other through an internal interface.

[0118] Processor 302 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0119] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras.

[0120] This application also provides a computer program product, including a computer program, which, when executed, is used to implement the above-described intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras.

[0121] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0122] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0123] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0124] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0125] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0126] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0127] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0128] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A method for intelligent monitoring of vegetation phenology in photovoltaic parks based on digital cameras, characterized in that, Includes the following steps: Acquire vegetation growth images of the target photovoltaic park and perform a preset semantic segmentation operation on the vegetation growth images to obtain the segmentation mask corresponding to the vegetation growth images. Based on the segmentation mask and the pre-built radiative transfer model, the pixel values ​​of different microenvironment regions in the vegetation growth image are subjected to a preset normalization correction operation to generate corresponding illumination correction images. Multi-channel pixel values ​​corresponding to each vegetation area in the illumination-corrected image are extracted, and the corresponding green chromaticity index sequence is calculated based on the multi-channel pixel values. The green chromaticity index sequence is then smoothed and denoised to extract the key phenological periods of each vegetation in the different microenvironment areas. Based on the key phenological periods, the phenological difference index corresponding to the target photovoltaic park is calculated to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index.

2. The intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras according to claim 1, characterized in that, The process of acquiring vegetation growth images of the target photovoltaic park and performing a preset semantic segmentation operation on the vegetation growth images to obtain a segmentation mask corresponding to the vegetation growth images includes: Based on the image acquisition requirements of the target photovoltaic park, the deployment optimization targets for the target digital cameras are determined. The array geometric parameters of the target photovoltaic park are obtained and input into a pre-constructed solar radiation transmission model. In conjunction with the deployment optimization target, spatial simulation calculations are performed to determine the optimal three-dimensional deployment position of the target digital camera, so that the target digital camera meets the preset area coverage requirements at the optimal three-dimensional deployment position. The target digital camera is deployed at the optimal three-dimensional deployment site, and the vegetation growth images of the target photovoltaic park are acquired through the target digital camera, wherein the target digital camera meets the preset weather resistance requirements of its internal components; At a preset edge computing node, based on a pre-built semantic segmentation model, a preset semantic segmentation operation is performed on the vegetation growth image to obtain multiple category labels corresponding to the vegetation growth image and classification masks corresponding to the multiple category labels. Based on the multiple category labels and the classification masks, the corresponding segmentation mask is determined. The multiple category labels include photovoltaic panel body category, under-panel shaded vegetation area category, inter-panel vegetation area category, peripheral full-sunlight vegetation area category, bare soil category, and sky category.

3. The intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras according to claim 2, characterized in that, The step of performing a preset normalization correction operation on the pixel values ​​of different microenvironment regions in the vegetation growth image based on the segmentation mask and the pre-constructed radiative transfer model to generate corresponding illumination-corrected images includes: At the edge computing node, the corresponding empirical model coefficients are determined according to the segmentation mask, and the theoretical total irradiance of the pixels corresponding to the different microenvironment regions generated by the radiative transfer model is obtained. Based on the theoretical total irradiance and the empirical model coefficients, a preset compensation operation is performed on the original pixel brightness values ​​corresponding to the different microenvironment regions to obtain target pixel values ​​that meet the preset vegetation color reflection requirements, so as to determine the illumination correction image based on the target pixel values.

4. The intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras according to claim 3, characterized in that, The process of extracting multi-channel pixel values ​​corresponding to each vegetation region in the illumination-corrected image, calculating the corresponding green chromaticity index sequence based on the multi-channel pixel values, and smoothing and denoising the green chromaticity index sequence to extract the key phenological periods of each vegetation in different microenvironment regions, and calculating the phenological difference index corresponding to the target photovoltaic park based on the key phenological periods, includes: In the edge computing node, the image matrix corresponding to the illumination correction image is converted into the corresponding floating-point data, and the corresponding multi-channel pixel value is extracted according to the floating-point data. Based on the multi-channel pixel value and the preset vegetation area mask, the green chromaticity index corresponding to each vegetation area is calculated, and the green chromaticity index sequence corresponding to multiple vegetation areas is determined through the green chromaticity index. The green chromaticity index sequence is sent to a preset cloud server, whereby the cloud server performs smoothing and noise reduction on the green chromaticity index sequence to obtain a corresponding noise-reduced sequence. Curve fitting is then performed on the noise-reduced sequence to generate a corresponding fitted sequence. Based on the fitted sequence, the key phenological periods of each vegetation in each vegetation region are extracted, wherein the key phenological periods include the vegetation greening period and the vegetation withering period. The phenological period difference between different locations of the photovoltaic panel is calculated based on the key phenological period, and the phenological difference index is calculated based on the phenological period difference. The different locations of the photovoltaic panel include the lower area of ​​the photovoltaic panel, the area between photovoltaic panels, and the outer area of ​​the photovoltaic panel.

5. The intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras according to claim 3, characterized in that, The mathematical expression for the target pixel value is: in, The original pixel brightness values ​​of the pixels corresponding to the different microenvironment regions; This represents the theoretical total irradiance of the pixels corresponding to the different microenvironment regions. These are the empirical model coefficients corresponding to the various category labels; This represents the target pixel value.

6. The intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras according to claim 4, characterized in that, The mathematical expression for the phenological difference index is: in, This indicates the accumulated days of the key phenological period corresponding to the area under the photovoltaic panel; This indicates the accumulated days of the key phenological period corresponding to the photovoltaic peripheral area; This indicates the accumulated days of the key phenological period corresponding to the inter-panel area; This represents the phenological difference index.

7. A smart monitoring device for vegetation phenology in a photovoltaic park based on a digital camera, characterized in that, include: The semantic segmentation module is used to acquire vegetation growth images of the target photovoltaic park and perform preset semantic segmentation operations on the vegetation growth images to obtain the segmentation mask corresponding to the vegetation growth images. The correction module is used to perform a preset normalization correction operation on the pixel values ​​of different microenvironment regions in the vegetation growth image based on the segmentation mask and the pre-built radiative transfer model, so as to generate the corresponding illumination correction image. The quantitative evaluation module is used to extract the multi-channel pixel values ​​corresponding to each vegetation area in the illumination-corrected image, calculate the corresponding green chromaticity index sequence based on the multi-channel pixel values, and perform smoothing and noise reduction processing on the green chromaticity index sequence to extract the key phenological periods of each vegetation in the different microenvironment areas. Based on the key phenological periods, the module calculates the phenological difference index corresponding to the target photovoltaic park to quantify the ecological effect of photovoltaic facilities in the target photovoltaic park through the phenological difference index.

8. An electronic device, characterized in that, include: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the intelligent monitoring method for vegetation phenology in photovoltaic parks based on a digital camera as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, The computer program is executed to implement the intelligent monitoring method for vegetation phenology in photovoltaic parks based on digital cameras as described in any one of claims 1-6.

Images