This paper approaches the problem of recognizing and localizing insulators with curved smooth and weak textured surfaces in a cluttered scene for the 220kv/330kv high-voltage live-line cleaning robot (HLCR). In this work we introduce SUSAN edge-based scale invariant feature (SESIF) to describe insulators, which is robust to noise, cluttered scene, changes on viewpoint and illumination. A coarse-to-fine method is applied to recognize and localize insulators using SESIF. Firstly, a coarse location is estimated with general Hough transform and RANSAC method. Secondly, an accurate location of insulator is found with an edge-based tracker. Results are demonstrated for a 220kv insulator in a cluttered scene
展开▼