Zhaowei Qu

Vision-Based Moving Objects Detection with Background Modeling

Video detection has become an efficient technique support for collecting parameters of urban traffic. Detection of moving objects with background model in complex environment is developed in this paper. 1) In order to obtain moving objects from the video sequence efficiently, a background initialization algorithm based on clustering classifier is presented, all stable non-overlapping intervals in the temporal training sequence of each pixel are located as possible backgrounds by slip window; then the background interval is obtained from the classified data set of possible backgrounds by unsupervised clustering. 2) According to spatial-temporal property of pixels, the paper also presents Mixture Gaussian background update algorithm based on object-level with moving segmentation. The method can get over the effect of object’s long-term stop. The proposed approach is validated under real traffic scenes. Experimental results show that moving objects detection is robust and adaptive, can be well applied in real-world.

Multi-level trajectory learning for traffic behavior detection and analysis

Motion patterns can be learnt automatically based on object trajectories data extracted by means of video tracking, which is an effective approach for modeling and analyzing traffic behavior. In this paper, a multi-level motion pattern learning approach for traffic behavior analysis is presented, which takes into account the spatial characteristics, direction characteristics, and type characteristics of trajectories. At the spatial level, improved Hausdorff distance measurement is applied to construct a spatial similarity matrix of the trajectories collected, and spectral clustering is used to achieve spatial pattern learning. At the directional level, the start and end points of trajectories are fitted using a Gaussian mixed model to extract the distribution of entry and exit zones. Then, the direction pattern is obtained from the regional centers of the pairwise distribution zones. At the type level, the type pattern is acquired through a K-means clustering algorithm that considers multiple classification features of trajectories. Based on the learned multi-level motion patterns, abnormal behavior detection algorithms are further developed by means of pattern matching. Finally, our approach is tested with several video sequences from real-world traffic scenarios. Some typical traffic behaviors in the test scenarios are successfully recognized and analyzed and examples of abnormal traffic behaviors are also reliably detected.

Research on Variable Lane Signalized Control Method

Taking the variable lane as a research object, the paper designs the detector layout plan according the traffic flow characteristics. Then, it determines the invariable lanes attribute according to the detector data and proposes the signalized control method of invariable lanes. In order to avoid the traffic flow confusion at approaches because of the drivers entering into the wrong traffic lane when the lane’s attribute is changed, paper lays out a fictitious stop line and proposed a pre-signal control method based on the fictitious stop line. The suggested strategy was verified with simulation software VISSIM. The results showed that this article method may enhance the road intersection traveling capability and reduces the delay by comparing with different channelization and control methods.

Nighttime Vehicle Detection Based on Illumination Visibility Recognition

Video detection is a valuable application in intelligence transportation systems. Some common traffic flow parameters like volume, velocity and vehicle category could be estimated automatically by video detection in real-time. At present, most of Video detection systems focus on algorithms designed and applied for day-time conditions. In this paper, a visual-based vehicle detection method for nighttime conditions is developed. Firstly, we consider there were two kinds of traffic scenes at nighttime: lamp scene and non-lamp scene. We calculate the illumination visibility of the region of interest (ROI) background image to divide traffic scenes into two predefined categories based on Support Vector Machine (SVM). Then, Different algorithms are used to complete the vehicle detection for two night scenes. At last, we test our approach with several video sequences from realistic traffic scenes including lamp scenes and non-lamp scenes. Experimental results show good performance of vehicle detection for nighttime conditions.

Abnormal Pedestrian Crossing Detection Based on Video Processing

In this paper, the authors developed a novel method to automatically detect abnormal pedestrian crossing behavior based on video processing. Firstly, crossing pedestrian trajectories are obtained from motion detection and tracking. For estimating the crosswalk zone, the authors use start points and end points to construct the start point set and the end point set. These two point sets are fitted by 2-dimension Gaussian Mixed Model (2-D GMM) to extract the distribution of entry and exit zones. The authors use a K-means algorithm to acquire the initial parameters of GMM, and use an Expectation-Maximization (EM) algorithm to optimize the parameters. Based on these steps, the crosswalk region can be estimated automatically with the predefined hypothesis. A pattern-matching algorithm based on Bayes classifier is presented for abnormal pedestrian crossing detection. The proposed approach has been implemented and tested at crosswalks in the real world. Experimental results show good performance of abnormal pedestrian crossing detection.