Feature extraction has always been a difficult problem in the classification performance of synthetic aperture radar automatic target recognition (SAR-ATR). It is very important to select discriminative features to train a classifier, which is a prerequisite. Inspired by the great success of convolutional neural network (CNN), we address the problem of SAR target classification by proposing a feature extraction method, which takes advantage of exploiting the extracted deep features from CNNs on SAR images to introduce more powerful discriminative features and robust representation ability for them. First, the pretrained VGG-S net is fine-tuned on moving and stationary target acquisition and recognition (MSTAR) public release database. Second, after a simple preprocessing is performed, the fine-tuned network is used as a fixed feature extractor to extract deep features from the processed SAR images. Third, the extracted deep features are fused by using a traditional concatenation and a discriminant correlation analysis algorithm. Finally, for target classification, K-nearest neighbors algorithm based on LogDet divergence-based metric learning triplet constraints is adopted as a baseline classifier. Experiments on MSTAR are conducted, and the classification accuracy results demonstrate that the proposed method outperforms the state-of-the-art methods.
Recently, multimodal biometric systems have received considerable research interest in many applications especially in the fields of security. Multimodal systems can increase the resistance to spoof attacks, provide more details and flexibility, and lead to better performance and lower error rate. In this paper, we present a multimodal biometric system based on face and ear, and propose how to exploit the extracted deep features from Convolutional Neural Networks (CNNs) on the face and ear images to introduce more powerful discriminative features and robust representation ability for them. First, the deep features for face and ear images are extracted based on VGG-M Net. Second, the extracted deep features are fused by using a traditional concatenation and a Discriminant Correlation Analysis (DCA) algorithm. Third, multiclass support vector machine is adopted for matching and classification. The experimental results show that the proposed multimodal system based on deep features is efficient and achieves a promising recognition rate up to 100 % by using face and ear. In addition, the results indicate that the fusion based on DCA is superior to traditional fusion.
Feature extraction plays a key role in the classification performance of synthetic aperture radar automatic target recognition (SAR-ATR). It is very crucial to choose appropriate features to train a classifier, which is prerequisite. Inspired by the great success of Bag-of-Visual-Words (BoVW), we address the problem of feature extraction by proposing a novel feature extraction method for SAR target classification. First, Gabor based features are adopted to extract features from the training SAR images. Second, a discriminative codebook is generated using K-means clustering algorithm. Third, after feature encoding by computing the closest Euclidian distance, the targets are represented by new robust bag of features. Finally, for target classification, support vector machine (SVM) is used as a baseline classifier. Experiments on Moving and Stationary Target Acquisition and Recognition (MSTAR) public release dataset are conducted, and the classification accuracy and time complexity results demonstrate that the proposed method outperforms the state-of-the-art methods.
Understanding a scene provided by Very High Resolution (VHR) satellite imagery has become a more and more challenging problem. In this paper, we propose a new method for scene classification based on different pre-trained Deep Features Learning Models (DFLMs). DFLMs are applied simultaneously to extract deep features from the VHR image scene, and then different basic operators are applied for features combination extracted with different pre-trained Convolutional Neural Networks (CNN) models. We conduct experiments on the public UC Merced benchmark dataset, which contains 21 different areal categories with sub-meter resolution. Experimental results demonstrate the effectiveness of the proposed method, as compared to several state-of-the-art methods.