本文
No.8(2013)3.Development of harmful birds recognition in video
Norihiro Ohira, Shinichi Tomiyama, Kazushi Miyatake, Yuka Hirai, Rieko Inuzuka
A new method for recognizing harmful birds in video has been developed. Some conventional methods recognize birds on power cables, branches or on the ground. However, from the perspective of preventing bird injuries, it is important to recognize birds before, rather than after, they land. Thus, we devoted our study to detecting birds before they land.
We focused on the characteristic that birds flap their wings during landing, and we developed a method to detect fluttering birds. A Hidden Markov Model (HMM) was used to improve recognition rate. In the HMM we correlated the state of regions with a bird’s state such as flapping or stopped and estimated the state and number of birds in the frame. We confirmed that this method has a 95% recognition rate via HMM and 70% without HMM.
Keywords
Video recognition, Harmful birds, Hidden markov model