An Approach of Animal Detection Based on Generalized Hough Transform

摘要

To detect animal objects under complicated background, a new approach is proposed to detect animal objects using shape feature. At first, a model is built by devising Canny hierarchical structure, then candidate fragments of the back and buttock are extracted using Generalizing Hough Transform, at last the matching task is accomplished based on restricting relationship of relative positions between the back and buttock. The experiment result shows that our approach achieves a promising detection rate under posture variations, scale variations and background clutter.