2016 · 10 citations · 17 references
ComputingComputer ScienceComputer VisionFeature EngineeringMachine LearningFeature DetectionHolographyObject DetectionObject RecognitionPattern RecognitionSupport Vector MachinesAutomatic Target RecognitionData AugmentationMillimeter-wave SecuritySparse Feature RepresentationSvmsTarget ClassificationTarget DetectionTarget Detection Training
Automatic Target Recognition (ATR) technology is of great significance in security inspection, while traditional object detection methods are proved not efficient in human body millimeter-wave images. In this paper, we propose a synthetic objection detection method for millimeter-wave images. We choose saliency, SIFT and HOG features to form image descriptors. According to sparse representation, the features are encoded again and fed to a linear SVM for target/non-target classification. Previous works proved that the amount of training samples would influence the efficiency of SVM classifiers. Thus, we utilize several simulating methods for data augmentation, aiming to increase the number of training samples before training linear SVM classifiers. The experimental results show that our approach is efficient in target detection of human body millimeter-wave images. Moreover, classifiers trained on larger sets with simulated samples have better performance in classification on our testing dataset.
17
Corinna Cortes, Vladimir Vapnik · Machine Learning · 1995 · 40.5K citations · Full text
Histograms of Oriented Gradients for Human Detection
Navneet Dalal, Bill Triggs · 2005 · 31.9K citations · Full text
A training algorithm for optimal margin classifiers
Bernhard E. Boser, Isabelle Guyon, Vladimir Vapnik · 1992 · 11.6K citations