Photovoltaic panel bracket image recognition and quantity calculation

How intelligent image processing technology can improve PV panel condition monitoring?

The research of this paper is to address this issue with the aid of intelligent image processing technology. In this study, an intelligent PV panel condition monitoring technique is developed using machine learning algorithms. It can rapidly process, analyze and classify the thermal images of PV panels collected from solar power plants.

How are PV panels inspected and photographed?

During the process of image collection, all four types of PV panels described in Sect. 2 are inspected and photographed when they are in operation. When taking the images, the camera was installed 0.6–1.0 m above the PV panels to simulate a scene where a drone carries the camera to monitor the PV panels.

How to evaluate PV panel extraction ability of PVI?

In order to evaluate the PV panel extraction ability of PVI more objectively and clearly, first, we calculated the PVI of all the images in the PVP dataset. Then, we transformed the PVI images into binary images using the Otsu [ 50] method. The evaluation metrics show that the mean values of IoU and F1 are 57.64% and 68.49%.

What is a multi-resolution dataset for PV panel segmentation?

This study built a multi-resolution dataset for PV panel segmentation, including PV08 from Gaofen-2 and Beijing-2 satellite images with a spatial resolution of 0.8 m, PV03 from aerial images with a spatial resolution of 0.3 m, and PV01 from UAV images with a spatial resolution of 0.1 m.

How can PV panels be detected and segmented?

PV panels can be detected and segmented from satellite or aerial images by designing representative features (e.g., color, spectrum, geometry, and texture).

How to extract PV panel information from a PvP dataset?

Wang et al. [ 17] trained their semantic segmentation model with the PVP dataset in the same year. Both studies demonstrated that accurate PV panels area can be extracted using red, green, and blue band images. Therefore, we used RGB band information to extract PV panel information.