Research On SAR Image Automatic Target Recognition Algorithm

Automatic Target Recognition technology is under the condition of without humanintervention, using SAR image collected data, accurate to find the region of interest, andextracting the characteristic information of the target class attribute, target identification. Inthe SAR ATR system mainly consists of three steps, i.e. SAR image preprocessing, thetarget feature extraction of SAR image and SAR image recognition. In this paper, threesteps, completed the following tasks:1. SAR image preprocessing: in this paper, research a SAR image denoising methodbased on complex contourlet transform and hidden Markov tree model. Using multi-scale,multi-directional and translational invariance features of complex contourlet transform, thismethod combined with HMT model can accurately describe the correlation of complexcontourlet domain coefficients between different scales. According to the quantitativeanalysis of experimental results, the method has achieved good denoising effect.2. Feature extraction method of SAR images: manifold learning method is a basicmethod in pattern recognition, Maximum Interclass Distance Embedding innovativemethods embodied in the combination of principal component analysis and the differencesin the characteristics of Local Preserving Projection neighborhood information, and at thesame time into the class information of the data set. By using this method is to find amapping of linear embedding and the SAR image features of different categories ofprojection was far away from each other.3. SAR image target recognition algorithm: This paper presented a SAR imagerecognition algorithm of neural network ensemble. The method solves the SAR imageazimuth sensitive issues, in order to realize target classification and target feature space totrain a neural network, and the recognition of multiple one-way target recognition resultsare used in combination with another two neural network, improve recognition accuracy.Automatic target recognition of SAR image of three kinds of method of these threestages and the application in the SAR image automatic target recognition, through thesimulation experiment quantitative analysis, target recognition has reached its final target recognition rate is higher than other more outstanding algorithm rate, the paper put forwardthe arithmetic is feasible.