Classification of vehicle type and make by combined features and random subspace ensemble

The identification of the make and model of vehicles from images captured by surveillance camera, also referred to as vehicle type recognition, is a challenging task in intelligent transportation system and automatic surveillance. In this paper, we first comparatively studied two feature extraction methods for image description, i.e., the MPEG-7 edge orientation histogram (EOH) and the pyramid histogram of oriented gradients (PHOGs). EOH captures the spatial distribution of edges by detecting five predefined types of edge directions. PHOG represents the local shape by a histogram of edge orientations computed for each image sub-region, quantised into a number of bins. Compared with previously proposed feature extraction approaches for vehicle recognition, EOH has the advantage of small feature size, economic calculation cost and relative good performance and PHOG has the ascendency in its description of more discriminating information. A composite feature description from PHOG and EOH can further increase the accuracy of classification by taking their complementary information. We then investigate the applicability of the random subspace (RS) ensemble method for vehicle classification based on the combined features. A base classifier is trained with a randomly sampled subset of the original feature set and the ensemble assigns a class label by majority voting. Experimental results using more than 600 images from 21 types show the effectiveness of the proposed approach. The composite feature is better than any single feature in the classification accuracy and the ensemble model produces better performance compared to any of the individual neural network base classifier. With moderate ensemble size 30, the random subspace ensembles offers a classification rate close to 94%, showing the promising potential in real applications.