An integrated method for multiple object detection and localization

Abstract

The objective of this paper is to use computer vision to detect and localize multiple object within an image in the presence of a cluttered background, substantial occlusion and significant scale changes. Our approach consists of first generating a set of hypotheses for each object using a generative model (pLSA) with a bag of visual words representing each image. Then, the discriminative part verifies each hypothesis using a multi-class SVM classifier with merging features that combines both spatial shape and color appearance of an object. In the post-processing stage, environmental context information is used to improve the performance of the system. A combination of features and context information are used to investigate the performance on our local database. The best performance is obtained using object-specific weighted merging features and the context information. Our approach overcomes the limitations of some state of the art methods. © 2008 Springer Berlin Heidelberg.