Target Recognition Algorithm Based On Gmm And Feature Weighted Svm Research

This paper investigates on the target recognition by TV seeker and the relative investigations have been conducted from the following three aspects.Part1:Based on analyzing the common noise reduction algorithms, the wavelet threshold denoising method has been mainly discussed. This paper analysis the selecting of the important parameters of the threshold function and the threshold value and it offers a self-adaptive threshold filtering algorithm. The numerical investigation denotes that this method could obtain higher peak signal-to-noise ratio, as well as better noise suppressing effect.Part2:For single frame image segmentation, This paper combines the Gauss mixed model (GMM) with maximum entropy method. The numerical experimental analysis shows that this method has a better effect compared with the histogram maximum entropy segmentation method. The moving target detection algorithm with mixture Gauss model is employed for the continuous image sequences. Because the results are vulnerable to light, random noise and shadow effects of the deficiency, this paper presents an improved algorithm. Each pixel intensity information has been used instead of the third chrominance. In the foreground detection, the random noise suppression mechanism is added. Results show that the improved algorithm not only could segment successful but also could remove the shadow and random noise.Part3:This paper also discusses the Hu invariant moments, affine invariant moments and contour feature extraction methods. Each weight coefficient for feature values have been calculated according to the contribution variation of characteristics for different classifiers. Existing algorithms take the importance of sample into account but neglect the relative importance of each feature with respect to the classification task. Feature weighted SVM algorithm is proposed to classification the target images. The numerical investigated also reviewed that the feature weighted support vector machine method is of higher recognition accuracy than the un-weighted.