Cross-arms Identification with Adaptive Digital Image Processing

This paper presents an adaptive digital image processing algorithm to improve the identification and classification process of wood cross-arms in the context of automated image inspection robot for distribution power lines. Processing of images that come from outdoor environment is a very difficult task due to the interference of variation of lightning conditions, especially over non uniform objects like wood cross-arms. Usual approaches take advantage of threshold and reference variation in accordance to the best condition to extract desired image object. The proposed adaptive algorithm is based on the Adaptive Digitized Straight Line Segments (ADSLS) that sum up the capability of dynamic change of image segmentation process like binarization and contour extraction. Proposed algorithm was tested on the identification and classification of different type of wood cross-arm with polyurethane protective treatment and results shown improvement in the image segmentation process.