Research Of Automatic Targets Acquisition Based On Interacting Multiple Model

Target acquisition has been widely used in the aviation, war industry, video monitoring and many other fields. Now target acquisition algorithms are mostly manual or semi-automatic, the capture process requires human intervention. When the target has a faster movement speed, manual mode cannot capture the target in real-time. Manual intervention cannot be used in some bad natural environment conditions. Research of Automatic Target Acquisition(ATA) can break through the bottleneck of the traditional capture methods; achieve the goal of capture the target whose motion is changeable. ATA contains two major steps, target catch and acquire. The main problem is how to quickly get the real target from a large number of suspected targets.In order to solve the above problem, this paper presents a dynamic model ATA method and a static model ATA method, the methods use the targets’ own characteristic to get the target. To reduce the memory consumption used to store the feature points and the matching computation, dynamic model ATA method mainly uses the estimation result of the interacting multiple model(IMM) to filter the output feature points of the optical method. The method artfully uses the target motion parameters to filter the suspect targets. Static model ATA extracts the edge feature to compute the regular targets’ length, width, canter information and the irregular targets’ centroid information. Then use the estimate results of the IMM to add new static feature models. And continue to introduce the updating static parameters to the acquisition process. Static model ATA method can adapt to the form change of the target.To test and verify the feasibility of the two methods, we use Matlab, Visual Studio software platform and the Open CV library to implement the above algorithms. When we use the dynamic model ATA method to acquire bicycles, motorcycles and pedestrians objects, the threshold of the method can be adjusted according to the different status of the target, then we use fewer feature points accurately acquire the targets. Static model ATA method continuously learns centroid characteristics of faces and cars, adapts to the form changes of the targets in the capture process, and simplifies the process of target acquisition.