Crow birds detection using HOG and CS-LBP

A robust image processing technique capable of detecting and localizing objects accurately plays an important role in many computer vision applications. In this paper, a feature based detector for birds is proposed. By combining Histogram of Oriented Gradients (HOG) and Center-Symmetric Local Binary Pattern (CS-LBP) as the feature set, detection of crows under various lighting conditions could be carried out. A dataset of crow birds with a wide range of poses and backgrounds was prepared and learned using linear Support Vector Machine (SVM). Experiments on different test images show that HOG and CS-LBP based descriptors can achieve 87% accuracy.