Infrared small target detection method based on generative adversarial network

Abstract

The invention discloses an infrared image small target detection method based on a generative adversarial network. According to the method, a false background image block is reconstructed for a targetimage block containing a small target; pairing the false background image block with the target image block to obtain first pairing data, training the generative adversarial network by using the data; pairing the two original background image blocks which do not contain the small target in the image sequence block to obtain second pairing data; continuously training by using the data to obtain atarget generative adversarial network model; and inputting a to-be-detected image block sequence into the target generative adversarial network model to obtain a detection sequence result, splicing the detection sequence result according to the size of the to-be-detected original infrared image, making a residual error between the spliced image and the to-be-detected original infrared image, and filtering the image after the residual error is made to obtain a detection result. Through the method provided by the invention, background interference can be greatly suppressed while a relatively high detection rate is ensured.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an infrared small target detection method based on a generative confrontation network. Background technique [0002] In recent years, infrared small target detection has been widely used in early warning systems as an indispensable part of infrared search and track systems. On the one hand, since the imaging distance of infrared images is usually tens to hundreds of kilometers, the signal is severely attenuated in the atmosphere, and the target energy is very weak on the image. Similarly, due to the imaging distance, the target usually occupies a small area and lacks texture and structure information; on the other hand, the background and noise occupy a large proportion in the infrared image, and the background is complex and constantly changing, resulting in the The signal-to-noise ratio in the image is low. Therefore, the infrared small target detection method has be...

AI Technical Summary

Problems solved by technology

Although the assumptions and principles of methods based on background suppression are relatively simple, the detection effect of such methods is not ideal

Images