Detection of moving objects in infrared image sequences using two-dimensional modified double channel motion detector network

Moving objects detection in a complicated surrounding is one of the most significant issues to be solved in the infrared target recognition because of the time-varying background. Biomimetic approach is a promising solution for tackling this problem. Visual detection and processing of motion in flies is thought to occur based on an elementary delay-and-correlate operation at an early stage in the visual pathway, which is the so-called `Reichardt detector´. However, the original Reichardt detector is a coarse description of the visual system of the fly. In this paper, we studied moving target detection method in a complicated background using a two-dimensional 2 quadrant motion detector network consisting of a grid of modified double channel motion detectors. At first, the scheme of modified double channel motion detector is analyzed and general properties of the modified double channel motion detectors are presented in detail. Then, the modified double channel motion detectors are interconnected in a grid to form a two-dimensional detectors network that can be used to model motion in two-dimensional scenes. The infrared image sequences including moving vehicle objects in the complicated background are used to confirm the proposed method in the experiment. The experimental results illustrate that, the proposed method achieves much better detection performance in infrared frames than the original Reichardt´s model. Moving object could be detected in infrared frames without the interference of background.