Small Animal Intrusion Detection Based on RAFT

Abstract

Due to the variety, size and shape of the invasive animals in substation, the general target detection and case segmentation tasks are often faced with the problems of fewer data and lack of diversity. Optical flow estimation can accurately detect the boundary of the moving object through the change of optical flow during the moving process of the object. Therefore, RAFT for small animal intrusion detection can not only effectively eliminate the interference of non-invasive objects, but also cancel the range of intrusion targets, and improve the generalization ability of moving object intrusion detection. The experimental results show that the detection effect is good for small invasive animals with low speeds.