Insect vision inspired object detection in the video with a moving cluttered background

Motion information is an important cue for insects to perceive the surrounding environment. Flying insects, demonstrate extraordinary capability in locating and detecting visual objects in a cluttered moving background. Utilizing insect vision in computational models is an intriguing challenge. We inherit the two-dimensional elementary motion detector model and improve the small-field model for video applications, which effectively suppresses the disturbance of the moving background directly in front of cameras. And then we propose an object detection method using peak, thresholding and direction voting, which locates the uncertain targets by detecting peaks in the modulus matrix, coupled with thresholding and then removing false targets according to the consistency of the argument matrix within the region of uncertain targets. We use the ObjectVideo Virtual Video visual-surveillance-simulation test bed to evaluate our detection results. Our methods produce robust target discrimination against the moving cluttered background.