Research on target type recognition algorithm of aerial infrared image

In order to improve the target type recognition rate of aerial infrared image under the new requirements of omnidirectional crossing and multitarget types, a recognition algorithm which has four steps is researched in this paper. Firstly the maximum between-cluster variance (Ostu) algorithm is applied to segment target from the infrared image. Secondly a new edge detection algorithm is proposed to get the target edge in the segmentation image. Thirdly the edge points are fitted to be a polygon and its Fourier descriptor is extracted to obtain the target feature. Finally the target type is identified by a classification recognition algorithm based on the BP neural network and the Fourier descriptor of target. The simulation results show that the recognition rate of the target type recognition algorithm is more than 80% for the typical aerial target on every complex condition, and the processing time is just 0.1s. So the researched algorithm can meet the requirements of real-time performance and high recognition rate.