A new vehicle type recognition mechanism based on image classification

作者：Zhang Yin^*; Liu Xun; Yang Zhixian; Lu Weiming; Zhang Sanyuan

来源：Journal of Computational Information Systems, 2013, 9(19): 7651-7658.

DOI：10.12733/jcis6814

• Histogram intersection kernels
• Image local features
• Machine learning classification
• Reconstruction error
• Sparse representation
• SVM
• Vehicle recognition
• Vehicle type recognition

Existing vehicle recognition algorithms require image's background to be simple or the same, which makes these algorithms to be poor scalable. By studying image local feature, image semantic model, and machine learning classification algorithms, we propose a new vehicle type recognition mechanism based on sparse representation and histogram intersection kernel. The new method can effectively reduce the reconstruction error by using sparse representation and tackle the big dimension problem by taking SVM classifier. In addition, we employ an online algorithm to learn the dictionary quickly. The experiment proves that the method can recognize vehicle types with a high accuracy in complex scenes.

• 出版日期2013
• 单位浙江大学