Research On Concealed Weapon Detection Underneath Person's Clothing Based On Image Fusion

Due to the color visual image has high resolution and natural color, and the infrared image can detect the weapon underneath the clothes, so, infrared and color visual image fusion can complement each other, and the fusion image not only can retain the concealed weapon information, but also can maintain the natural color and high resolution of the color visual image. Nowadays, most of the research on CWD(Concealed weapon detection, CWD)fusion detection are based on gray level images, and the research based on the color image fusion detection is very rare, and the research on the protection of privacy of the person is also very little. So, in this paper the concealed weapon detection algorithm based on color image and infrared image fusion is studied.The main studies contents of this paper are as follows:(1) Based on image segmentation, a new method of hidden weapon detection based on fuzzy C means clustering is proposed, which is based on FCM(fuzzy C- means Clustering)theory and mathematical morphology. Firstly, the color visual image is transformed to HSI color space, and the infrared image is processed with FCM and the segmentation image is obtained. Secondly, the segmentation image is processed with mathematical morphology of fill the holes, then the binary image only with weapon is obtained by the subtraction of those two images. Thirdly, gray fusion image is obtained by mosaic the weapon in the binary image to the I component. At last, the grayscale fused image is regarded as the new I component to be transform to the RGB color space, and the color image fusion is obtained. Experimental results show that the proposed method not only can obtain the true color of the source image and the clear target weapon, but also protect the privacy of the human body, and this method can achieve a better fusion effect.(2) Based on multi-resolution mosaic technology, the I component and the binary image are decomposed by NSST(Non-Subsampled Shearlet Transform, NSST), and the corresponding low frequency and high frequency components are obtained. The someinformation in the low frequency and high frequency add with different weights, and the fused low frequency coefficients and high frequency coefficients are obtained; then the gray fused image is obtained by NSST inverse transform. By comparing with the second chapter algorithms, the experimental results show that this chapter method not only maintain the all advantages of the second chapter, but also the weapon looks more natural in the fusion image and is close to the nature of the weapon in the source infrared image.(3) Two methods are proposed in this chapter for the image with special background. The main idea of this method is make the process of dim for the image with the high light background, which makes the background and the weapon be distinguished. The first method is that the weapon is dimmed firstly, then the process of mosaic is done, and the purpose of distinguish weapon from image is achieved. The first step of the second method is that the I component and binary image is transformed by NSST. The dimed weapon is obtained by adjusting the fusion weighted coefficient of high-frequency and low frequency. Experimental results show that the two methods have achieved the purpose of outstanding weapons, but the fusion image obtained by the second method looks more natural.