Research On Multi-class Object Detection Based On Object Proposals

Object detection is a hot topic in the area of computer vision,it is also one of the most basic and critical technology in intelligent video surveillance.Two general approaches for object detection have emerged.The former method requires that objects of interest are moving while the background is constant.The latter method tries to model foreground and learn features from the training sample which can establish the model.The well known “sliding window” paradigm is the most successful approaches to object detection.Sliding window classifiers scale linearly with the number of windows tested,and while single-scale detection requires classifying about millions windows per image,the number of windows grows by an order of magnitude for multi-scale detection.In order to solve the above problems,this thesis mainly focuses on multi-class object detection based on detection proposals.The research content of the specific as follows:(1)This thesis investigates popular detection proposals algorithm thoroughly and has a review on popular detection proposals algorithms and the corresponding improved algorithms.(2)This thesis proposes an effective and efficient proposal generation method which uses a Multi-layer and Multi-size Superpixel Segmentation scheme for object detection in infrared image.SLIC is applied to partition an infrared image into multi-layer and multi-size superpixels.Only the individual superpixel and the merging of two adjacent superpixels are used to create the candidate pool of object proposals.A superpixel-based center-surround feature is then defined to measure the discrepancy between the region of the proposal and its surrounding background.To evaluate the performance of the proposed method of proposal generation method,this thesis creates an Infrared Interested Object Image Dataset,and the ground-truth of the interested object segmentation is manually labeled.Compared with several state-of-the-art methods of proposal generation on IIOID,the MMSS-based method has overwhelming superiority in detection recall and is convenient for computation.(3)This thesis introduces the framework of multi-class object detection based on detection proposals.Firstly,objestness algorithm is used to generate a series of detection proposals of imput images.Then a multi-class classifier is used to evaluate these proposals,and the result of multi-target detection is obtained.The experimental results show that this method can accurately detect multi-class objects simultaneously.