Animal behavior recognition method based on multi-clue fusion

• Main Authors: SHEN LONGFENG, LIU HUAIYU, JIAN YULEI, GE FANGZHEN, HONG LIURONG, GAO XIANGJUN, LI XIANG
• Format: Patent
• Language: Chinese; English
• Published: 12.07.2022
• Subjects: CALCULATING; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides an animal behavior recognition method based on multi-clue fusion. The animal behavior recognition method comprises the following steps: constructing a behavior data set; extracting an optical flow image of animal motion from the RGB frame; inputting the RGB frames and the optical flow frames into an animal behavior recognition model constructed by a multi-clue fusion network to recognize animal behaviors, wherein the method comprises the following steps: performing feature extraction on the RGB frames and the optical flow frames by using a double-flow network to obtain apparent features and motion features of the animal behaviors; carrying out multi-clue feature exchange on the features, and discarding a feature map which loses influence on a final result; inputting the double-flow features of the MFSM module into an AFFM module, and performing adaptive feature fusion on the apparent features and the optical flow features to obtain motion information enhanced apparent features; and inpu