研究目的
To compare several methods for detecting failed solar panels and determine the most efficient method, focusing on drone detection and automatic detection using deep learning.
研究成果
Drone inspection combined with SSD for automatic failure detection is efficient, achieving an mAP of 49.11%. However, improvements are needed in learning data and orthographic image synthesis to reduce errors and enhance detection accuracy.
研究不足
The study faced challenges in detecting panel failures due to insufficient learning data. Additionally, errors occurred in the synthesized orthographic images, and manual adjustments were needed for accurate detection.
1：Experimental Design and Method Selection:
The study compared drone inspection with conventional inspection methods for detecting failed solar panels. Deep learning techniques, specifically SSD, were employed for automatic detection.
2：Sample Selection and Data Sources:
Thermographic images of a 4-MW solar power plant were used, including images of cluster failures, panel failures, and string disconnections.
3：List of Experimental Equipment and Materials:
DJI INSPIRE 1 (drone) and DJI ZENMUSE-XT (thermographic camera) were used for capturing images.
4：Experimental Procedures and Operational Workflow:
The process involved capturing thermographic images with a drone, detecting failures using SSD, and mapping the failures using OpenDroneMap.
5：Data Analysis Methods:
The effectiveness of SSD was evaluated using the mean average precision (mAP) index.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
