Peixun Liu

A lane marking detection and tracking algorithm based on sub-regions

In this paper, we present a lane marking detection and tracking method for the Lane Departure Warning system (LDWs). The method can both detect straight and curve lanes on the highway and urban road. The lane marking detection part is divided into three phases. First, bisect the ROI (region of interest). Then get lines by processing Hough transformation on each sub-region, after this phase, we can get a set of seed points. Finally, judge the lanes type (straight or curve) by its slope. If its a curve line, fitting the lane by using the rational cubic B-spline curves. In the lane tracking part, we use Extended Kalman Filter to make sure continuous lane detection results. In the experiment part, we use a dataset which includes virtual videos and real videos to measure our methods performance under the virtual and real environment. Experiment results show that the average detection rate of virtual video is 80%, and the real video is 85%. And the average processing time of virtual video is about 5.3ms/frame, and the real video is about 14.0ms/frame.

Early Flame Detection in Video Sequences based on D-S Evidence Theory

A multi-feature fusion early flame detection algorithm based on D-S evidence theory is proposed. In this algorithm, first the method based on YcbCr and RGB color spaces is used for extracting the flame region of interest. Then flame classifiers based on flame flicker frequency and flame image correlation between frames are selected as two features of D-S feature fusion, the basic probability assignment functions of two features are defined. Finally, the combination discipline of D-S evidence theory is used to determine the final result of all the feature classifiers. Experiments show that the proposed detection algorithm gives 89.5% correct flame rate with a 4.5% false alarm rate and the method is efficient and fast with wide application prospects in fire detection.

Improved SDA based on mixed weighted Mahalanobis distance

We propose an improved subclass discriminant analysis based on the mixed weighted Mahalanobis distance, with the aim of resolving the classification problem that samples are multi-subclass distributed and the computed vectors are sub-optimal. There are three contributions in this paper. First, we used some theoretical support to improve the traditional discriminant criterion function in subclass discriminant analysis. Second, we propose a new distance measure called the mixed weighted Mahalanobis distance (MWMD), which considers the influence of the sample size and sample scatter. Finally, inspired by approximate weighted linear discriminant analysis, we applied MWMD to the improved criterion function. Our experimental results confirmed that the proposed method has a better classification performance than other discriminant analysis methods, using an artificial dataset and some benchmark databases.

An Interactive Segmentation Method of LiDAR Data

In order to alleviate the problems inherent of automatic segmentation of LiDAR data, an interactive graph-cut segmentation method of LiDAR data is proposed. Firstly, the research background and the basic conceptions of the interactive graph-cut algorithm are introduced. Secondly, by analyzing the characteristics of LiDAR data, four-dimensional feature vectors are extracted, which as the graph-cut algorithms input. Thirdly, the optimal parameter is estimated according to a new Sample-fitting method. At last, the experimental results show that this interactive segmentation method of LiDAR data is able to accurately locate the buildings region with less interaction, and at the same time guarantee the accuracy rata when buildings and trees are connected to each other.

Multifeature Fusion Vehicle Detection Algorithm Based on Choquet Integral

Vision-based multivehicle detection plays an important role in Forward Collision Warning Systems (FCWS) and Blind Spot Detection Systems (BSDS). The performance of these systems depends on the real-time capability, accuracy, and robustness of vehicle detection methods. To improve the accuracy of vehicle detection algorithm, we propose a multifeature fusion vehicle detection algorithm based on Choquet integral. This algorithm divides the vehicle detection problem into two phases: feature similarity measure and multifeature fusion. In the feature similarity measure phase, we first propose a taillight-based vehicle detection method, and then vehicle taillight feature similarity measure is defined. Second, combining with the definition of Choquet integral, the vehicle symmetry similarity measure and the HOG