Research On Infrared Target Detection Technology Under Various Imaging Distances

In recent years,with the development of digital image processing technology and the continuous improvement of the quality of infrared detectors,target detection technology based on infrared imaging is widely applied in the military field.The traditional search and locating of enemy fire points are mainly realized by radar,and that remains some problems during the war including easy to expose,the radar volume is too large and etc.In this paper,photoelectric technology and related infrared image processing algorithm are used to detect the muzzle flame of mortars and other weapons.Infrared images are easily subject to noise and interference under complicated environments,which brings great difficulties to infrared target detection.Therefore,it is of great significance to detect such special targets in a low contrast and signal-to-noise ratio infrared image.This paper focuses on the detection of muzzle flame targets under various imaging distances,which could mainly study from the following aspects:For the infrared image noise and difficulties of infrared image detection,several components of images including the principle and characteristics of infrared imaging system are studied for providing theoretical basis.In terms of noise suppression of images,the method of median filtering and weighted median filtering is first studied and analyzed.For the blind pixel problem in infrared images,a gradient-based blind pixel detection and correction method is proposed.Experiments show that the method can effectively remove the blind pixel noise in the image.About the low contrast of infrared images,two methods of image enhancement such as histogram equalization and homomorphic filtering are studied.For the situation where the infrared camera is close to the target,combined with the target characteristics of muzzle flame,a target detection algorithm based on Vibe and target characteristics is proposed.The initialization and update rate of the background model are adjusted,and false detected targets are eliminated according to the target features.Finally,the algorithm is implemented on embedded Linux platform(NVIDIA Jetson TX2)and compared with other traditional algorithms.Tests on infrared video sequences under different conditions show that proposed algorithm is effective for muzzle flame target.For the circumstances where the infrared detector is far from the muzzle flame,the target takes up less pixels in the image and lacks of characteristics of texture,color,shape.A new infrared small target detection algorithm based on NSST and phase spectrum of quaternion Fourier transform is proposed in this section.Firstly,the original image is decomposed at several scale and direction by NSST,and the high-frequency sub-band images in different direction of the image are obtained.Then,the phase spectrum of Fourier transform is introduced and extended to four data channels.For the reconstruction of image,phase spectrum of quaternion Fourier transform is utilized to combine the decomposed high-frequency sub-images.In the end,the reconstructed image is segmented by an adaptive threshold to accomplish the detection of small target in the infrared image.The superiority and robustness of proposed algorithm are confirmed by experimental analysis and comparison.