研究目的
To develop a computer vision method for soiling recognition in photovoltaic modules to enhance the efficiency and autonomy of cleaning robots.
研究成果
The developed method can effectively identify and quantify soiling on photovoltaic panels, enhancing the intelligence of cleaning robots. The algorithm was patented in Brazil, indicating its potential for practical application in solar energy generation.
研究不足
The region growing technique had difficulties with non-uniform dirtiness distribution, stopping when another region is achieved. The Hough method showed some false positives given by cell wiring and reflection in clean panels.
1：Experimental Design and Method Selection:
The study extends classic computer vision algorithms such as Region Growing and the Hough transform, incorporating pre-processing techniques based on Top Hat and Edge detection filters.
2：Sample Selection and Data Sources:
An image database was created with varying levels of wet sand application over photovoltaic solar panels to simulate uniform and non-uniform dirtiness.
3：List of Experimental Equipment and Materials:
Photovoltaic solar panels, wet sand, stones, and a cartesian robot for cleaning.
4：Experimental Procedures and Operational Workflow:
The method involves applying region growing and Hough transform algorithms to identify and quantify soiling on panels, followed by validation through experiments.
5：Data Analysis Methods:
The analysis includes histogram and mean analysis to correlate dirtiness levels with histograms amplitudes and means.
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
