A High Performance Imaging and Target Discrimination System Using SAR and ISAR

We propose a high performance, automated, and novel framework to behind-the-wall imaging and target discrimination. Our system has two paths: 1) stationary target detection using SAR images; 2) moving target detection using ISAR images. For the SAR imaging part, we first propose to apply high performance algorithm to form SAR images. Second, since feature extraction is the key component for a successful target discrimination system, we propose 4 different classes of features: 1) invariant moments; 2) low order Gaussian features based on Principal Component Analysis (PCA); 3) high order non-Gaussian features based on Independent Component Analysis (ICA); 4) shape features based on Fourier descriptors. Third, after features are extracted, we propose to apply proven and efficient classifier to classify the different targets. The robust classifier we use is called Support Vector Machine (SVM) that has several advantages, including no over training problem, global optimal solution, and computational efficiency. In the ISAR part, we have ISAR image formation algorithm, followed by a chirplet separation algorithm to separate different motions. After ISAR images are formed, the feature extraction and classification algorithms mentioned earlier will be applied for target discrimination. All the images and classification results will be displayed on a monitor.