Local feature-based multi-object recognition scheme for surveillance

In this paper, we propose an efficient multi-object recognition scheme for surveillance based on interest points of objects and their feature descriptors. In this scheme, we first define a set of object types of interest and collect their sample images. For each sample image, we detect interest points and construct their feature descriptors using SURF. Next, we perform a statistical analysis of the local features to select representative points among them. Intuitively, the representative points of an object are the interest points that best characterize the object. Finally, we calculate thresholds of each object for object recognition. User query is processed in a similar way. A given query image's local feature descriptors are extracted and then compared with the representative points of objects in the database. Especially, to reduce the number of comparisons required, we propose a method for merging descriptors of similar representative points into a single descriptor. This descriptor is different from typical SURF descriptor in that each element represents not a single value but a range. By using this merged descriptor, we can calculate the similarity between input image descriptor and multiple descriptors in database efficiently. In addition, since our scheme treats all the objects independently, it can recognize multiple objects simultaneously.