Multi-view Object Detection By Classifier Design And Interpolation

The detection of moving object is a very active branch in the image processing and computer vision field. Research in this domain involves wide application, including security surveillance, human-computer interface, the details analysis of human movement, etc. Based on the summary and analysis of the relevant research works home and abroad, we propose a novel framework for multi-view object detection based on classifier design and interpolation.First of all, we embed human knowledge in learning process, letting human operator to manually design a classifier for each view of the object. To make the template more reasonable, designer can adjust the template not only according to the current displayed image, but also his knowledge about the appearance of the object type. This is a manually designing process, and the resultant classifier will not depend heavily on the data.Given a set of training examples at different views, we select examples at a few key views and design one classifier for each of them manually. Then classifiers for more intermediate views can be interpolated from key views. The interpolation is conducted on the weights and positions of features, under the assumption that they can all be expressed as functions of view angle. Finally, the designed and interpolated classifiers are combined into a boosting framework to construct a multi-view classifier for multi-view object detection.Experiments of designed and interpolated single view classifier and combined multi-view classifier are conducted on pedestrian and car data sets and their performances are compared to corresponding learned classifiers. The results illustrate that the designed and interpolated classifiers give comparable performance as classifiers learned from data, and that the combined classifiers improve the whole performance just like their learned counterparts. The results of combined classifiers also further validate the effectiveness of the design and interpolation.