The background suppression algorithm based on the two-dimensional velocity vector histogram and the estimated risk

The small-target detection in infrared image is a difficult task in remote sensing fields. First of all, the reason of the increased false alarm rate under the cloud cluster background is analyzed. In order to avoid this kind of the false alarm, the background suppression algorithm based on the two-dimensional velocity vector histogram and the estimated risk is presented. The image sequence is filtered firstly under the higher false alarm rate to extract the further more interference points in the cloud region. Then the velocity vectors of all the detected points including the interference points are computed by the means of date association. And the two-dimensional velocity vector histogram is calculated with the velocity vector of all the detected points. It is found that the most of the detected points are the interference points in the cloud and the velocity and the direction of the cloud are identical in a certain field of view for some time. According to the characteristic of the cloud, the range of the cloud velocity vector is obtained based on the velocity vector histogram by means of the statistics. The false alarm points in the motional cloud are filtered out according to the velocity range. But a few false alarm points in the cloud may also exit. So the concept of estimated risk is presented to evaluate the possibility of false alarm point. The threshold of each part of the image is adaptively adjusted based the estimated risk evaluated by the complex level of background. Then the false alarm points in the complex cloud are filtered out based the threshold. The experimental results with the real image show that the proposed method can reduce the false alarm of the target detection under the cloud background and detect target successfully.