[show abstract][hide abstract] ABSTRACT: This paper presents a real-time vision-based vehicle detection system employing an online boosting algorithm. It is an online AdaBoost approach for a cascade of strong classifiers instead of a single strong classifier. Most existing cascades of classifiers must be trained offline and cannot effectively be updated when online tuning is required. The idea is to develop a cascade of strong classifiers for vehicle detection that is capable of being online trained in response to changing traffic environments. To make the online algorithm tractable, the proposed system must efficiently tune parameters based on incoming images and up-to-date performance of each weak classifier. The proposed online boosting method can improve system adaptability and accuracy to deal with novel types of vehicles and unfamiliar environments, whereas existing offline methods rely much more on extensive training processes to reach comparable results and cannot further be updated online. Our approach has been successfully validated in real traffic environments by performing experiments with an onboard charge-coupled-device camera in a roadway vehicle.
IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics) 07/2010; · 3.24 Impact Factor
[show abstract][hide abstract] ABSTRACT: This paper presents a real-time vision-based side vehicle detection system employing a parts-based boosting algorithm. Working at the part level, as opposed to the whole object level, enables a more flexible class representation and allows scenes in which the query object is significantly occluded to be classified. Therefore, a parts-based learning approach is proposed in order to better deal with side vehicle variability, illumination conditions, partial occlusions, and rotations. Most existing boosting learning algorithms usually select weak classifiers by minimizing a cost directly associated with the error rate, where the learned strong classifier may be sub-optimal for applications in terms of error rate. Nevertheless, the proposed Adaboost approach selects weak classifiers by minimizing multiple types of error functions. The idea is to define multiple types of error functions based on current strong classifier and each selected weak classifier results to represent different effects of each weak classifier. Therefore, weak classifiers can be selected with different requirements at the same time to avoid a sub-optimal solution. To reduce system computation, window-based tracking is employed. Moreover, Kalman filtering is used to predict the position of each part of vehicles in the image plane to effectively relocate the tracking windows. Compared with existing approaches, the proposed system appears to be capable of improving system efficiency and accuracy under varying lighting conditions, changing vehicle poses, and in the presence of partial occlusions. Our approach has been successfully validated in real traffic environments by performing experiments with a CCD camera mounted onboard a highway vehicle.
Systems, Man and Cybernetics, 2008. SMC 2008. IEEE International Conference on; 11/2008
[show abstract][hide abstract] ABSTRACT: This paper presents a multi-class boosting algorithm employing color-based Haar-like features. Traditional multi-class boosting algorithms basically regard multi-class problems as extensions of two-class problems. In particular, additional strong classifiers must be parallelly extended once the number of target classes increases. The idea in the proposed approach is to develop a single strong classifier which is capable of resolving multi-class problems. To make the multi-class algorithm tractable, the proposed system is required to select a set of weak classifiers which could classify multiple types of targets correctly. In contrast to standard Haar-like features that compute feature values based on gray level images, the seemingly novel Haar-like features require computation based on color images. Since the mapping from color image space to gray level image space is an epimorphism, detection algorithms using standard Haarlike features inevitably disregard color information available in original color images. Strong classifiers adopting the proposed color-based Haar-like features typically appear to have comparable performance, in the aspects of detection and correct classification rates, with fewer weak classifiers when compared with the one employing standard Haar-like features. The proposed boosting algorithm can improve system efficiency and resolve multi-class problems by a single strong classifier, whereas existing approaches are more complicated and the number of two-class classifiers could be relatively large. Our approach has been successfully validated in real traffic environments by performing experiments with a CCD camera mounted onboard a highway vehicle, where the targets are defined as passenger cars and motorcycles.
Signal-Image Technologies and Internet-Based System, 2007. SITIS '07. Third International IEEE Conference on; 01/2008
[show abstract][hide abstract] ABSTRACT: In this paper, a mobile robotic manipulation system equipped with an active eye-to-hand binocular vision system is proposed. The active vision system is able to observe arm gesture of a user by visually recognizing features on the arm and reconstructing the arm pose in Cartesian space. The target that the user wants to point at can thus be identified based on the arm pose. The user is then prompted with the image of the target on screen for confirmation by hand gesture. If confirmed, a vision- based control law is able to drive the robot toward the target position and to fetch the target by the robotic arm. Similarly, the target can be delivered to the user by real-time visual servoing. In fact, the proposed system can successfully interact with users by using binocular vision. Specifically, this robotic system could effectively help users to fetch targets that users designate by simple arm gesture.
Industrial Electronics Society, 2007. IECON 2007. 33rd Annual Conference of the IEEE; 12/2007
[show abstract][hide abstract] ABSTRACT: Due to the fact that existing surveillance systems usually employ fixed or automatically-patrolled cameras, they cannot effectively track invaders in real-time. Therefore, an automatic head tracking system in indoor cluttered environment is proposed. This system consists of two parts, automatic head tracking system and remote monitoring and control system. In the automatic head tracking system, the position of human head on image plane can be determined based on head tracking strategy. According to the position and size of human head on image plane, a suitable fuzzy rule is designed to maintain the position of human head near image center with appropriate size by controlling the active camera via pan, tilt, and zoom. Specifically, the position of the moving object on image plane can be determined using moving object detection method. Then a sizable tracking window to track human head is used to reduce redundant computation. Moreover, color recognition, template matching, and head contour matching methods are integrated for tracking human head robustly. In order to achieve better performance, Kalman filtering is employed to predict the position of human head on image plane. The proposed active head tracking system has been successfully validated in indoor cluttered environment.
Systems, Man and Cybernetics, 2006. SMC '06. IEEE International Conference on; 11/2006