Research On Infrared Small Target Detection Under Various Complex Backgrounds

Infrared technology occupies a very important position in the military and civilian fields.Various governments and large scientific research institutions are committed to the research of the hot research topic.Now the war requires the weapon system to have the ability to detect targets from a long distance.Target detection adds a lot of difficulty when the target is far away from the detector.At present,scientific research institutions have proposed many different target detection algorithms successively for infrared target images in different backgrounds.These algorithms are mainly based on high-pass or low-pass and morphological filtering.The above algorithms generally only have a remarkable detection rate for a specific background,so the infrared target detection algorithm that is universally applicable to various backgrounds is still a research difficulty.This article is devoted to the research of infrared small target detection algorithms that are generally applicable to various complex backgrounds,two infrared small target detection algorithms based on PFT and SVDD are proposed.In order to improve the target detection probability at the same time overcome the shortcomings of singlechannel image,we adopted the notion of multi-channel images.Then,the multi-channel images are synthesized by image fusion to form the final significant image.In order to overcome the shortcomings of the PFT algorithm that may cause missed detection in the target area,we consider the target detection problem as a class of classification problems.This paper focuses on the possibility of using the SVDD classifier to train infrared targets and verify it detection performance.In addition,how to narrow down the search area of target detection is also the focus of this study.The saliency map acquired by the PFT algorithm only includes the suspicious area where the target may exist,greatly narrowing down the scope of the target area to be searched for,saving the computation space.In order to better restore the size of the real target,this paper maps the circumscribed rectangle of the suspicious area back to the original image,and specifies the size of the test sample as 2 times the circumscribed rectangle of the suspicious area.Then use the same proportion of window scaling to make the final test sample the same size as the training sample.Then passing determine whether the test image is inside the hypersphere to achieve target detection in the suspicious area.In general,this article has the following innovations:1.In the PFT algorithm,the problem of small target detection is actually transform to a significant area detection problem.The PFT algorithm can well reflect the position of the image change.Therefore,this algorithm uses this algorithm to obtain the significant region of the infrared image.2.This article innovatively treats the target detection problem as a class of classification problems,and proposes an infrared small target detection algorithm based on SVDD.Using the SVDD algorithm to train the target sample of the simulation,a hypersphere containing as many samples as possible is obtained,and finally the target is detected by determining whether the sample to be detected is inside the hypersphere.This method has excellent effects and high detection accuracy.3.In order to narrow down the range of target detection and improve the detection accuracy of the target,this paper firstly uses PFT algorithm to obtain the significant region of the infrared image,The adaptively window is scaled to obtain the detection sub-image,and then the SVDD algorithm is used to determine the sub-image to achieve the final detection of the target.Therefore,using the SVDD algorithm to detect targets in the region of significance has better detection performance.