Target Detection And Tracking Technology Research

In this dissertation, a study on the problem of face detection and tracking is presented. As a key technique in the field of computer vision, face detection and tracking has many potential important applications in a wide range of fields including video surveillance, face recognition systems, content-based retrieval, advanced human and computer interaction.Techniques for addressing the object detection problem include those matching a two- and three-dimensional geometric model to images, and those using a collection of two-dimensional images of the object for matching. This dissertation will show that the latter view-based approach can be effectively implemented, allowing the detection of upright, tilted, and non-frontal faces in cluttered images. Under this framework, a method efface sample collection is presented, and then large amounts of relevant data are collected and classified. Based on this work, we carried out the research work:To deal with the essential problem of face detection in gray scale image, an algorithm based on subspace is presented. Because of the high complexity problem involved in the pure statistical approaches due to enormous feature space and the difficulty in collecting the "non-face" samples for training, mosaic image analysis is first adopted to filter and constrain the pattern classification problem in a greatly downsize subspace, thus facilitating the training procedure. In this dissertation, the performance of the mosaic image analysis is evaluated. The experiment shows that the mosaic image analysis can be used for face detection under some constrained conditions. For face detection in cluttered images, a Support Vector Machine based face detection algorithm trained under the mosaic image analysis constrained subspace is proposed. Experiment results demonstrate the effectiveness and robustness of the proposed method.What is more, we implement object-tracking algorithm combining optical flow and Pyramid representation of image. The results of the experiment show that this method can handle the problem of large vectu- movement and improve llie robustnessof tracking.