Hybrid C2 features and spectral residual approach to object recognition

To establish a fast and accurate visual object recognition system is a fundamental and formidable task. In most cases, rapidity and stability can be inconsistent. In order to solve this problem, a novel object recognition system, which is based on hybrid C2 features and saliency detection, is proposed in this paper. Spectrum residual method is introduced for fast location of object regions, and the C2 features are the main characteristic used for object distinction. After the candidate regions of possible target are extracted, the object recognition task is converted to a categorization problem, which can be solved by SVM (Support Vector Machine) classifier. To prove the feasibility and effectiveness of our system, several experiments are conducted, and the experimental results further illustrate the rapidity with high accuracy of our proposed recognition system.