Long-range Small Target Detection In Complex Background Using Multi-sensor Images

With the development of digital battlefield technology, the need for automatic target recognization(ATR) is urgent. The target detection technology, especially the long-range small moving target detection technology in complex environment, is very important in automatic target recognization, for the performance of the whole ATR system relies mainly on the target detection. The long-range small moving target detection technology can be applied not only to the battlefield, but also to some sensitive areas like the airport, and so on.As for the detection of weak point target with single infrared sensor, however, there are some drawbacks in the target detection process with one single sensor, for only one kind of information can be obtained from one single sensor. As a result, when the temperature of the target is lower than the background or the background is very complex, the target can be detected with very low probability.Aiming at the above-mentioned issues, this thesis has probed into the problem of multi-sensor long-range small moving target detection in complex environment by using the feature fusion and decision fusion in the information fusion theory. At first, this paper adopts the feature fusion detection method based on the Fisher classifier. Wherein, for infrared images, the features based on the joint histogram are proposed; while for vision images, this paper proposes Gaussian energy based feature. However, the targets can not be distinguished from the background when the targets'temperature is lower than the background's using the joint histogram features in infrared images. As a result, another method for getting the features is introduced, which is based on dual-rectangle window. At last, for the decision fusion, after getting the fuzzy decision-making of each sensor, fuzzy consistency theory is introduced to find the last decision result. Algorithm results are given based on different actual data.This paper has provided a satisfied solution to the problem of multi-sensor long-range small moving target detection. A lot of experiment results on different actual data indicate that this algorithm is robust and efficient for the long-range small moving target detection in complex environment.