Pattern Clustering Based Road Detection And Hardware Implementation Of Data Interfaces

Road Detection is the key technology of intelligent vehicles. Intelligent vehicles can be applied in many fields. They can transport materials without human supervisory in the hazardous area. Human beings may be killed or badly injured in this area such as the battle field and the industrial park. In daily life, intelligent vehicles can forestall traffic accidents and enhance safety of the road transport. A pattern clustering based road detection method is presented in this paper to detect road area in structured and semi-structured roads. Clustering collects data with similar features into one class. Fuzzy C-means and image segmentation are two methods of clustering methods.Structured roads such as national roads have distinct lane markings. Intelligent vehicles run fast on national roads. The processing time of road detection method is required to be as short as possible. A fuzzy C-means clustering based lane markings detection method is proposed in this paper to meet the requirement of the real-time video processing. The connected domain is used in this method to identify lane markings in edge images. The new method presented in this paper can detect lane markings from an image in less than 35 ms, which means the method can handle 25 frames in 1 second. The processing capacity satisfies the speed requirement of intelligent vehicles. In contrast to structured roads, the semi-structured roads do not have clear lane markings. But the road surface which is made of cement or bitumen is definitely different from the surrounding environment. An image segmentation based road detection method is presented to detect road area in semi-structured roads. The road area which is computed by the image segmentation method is used to calculate road curvature and borders. A novel curvature determined method which is based on geometry is proposed in this paper to select the most matched curvature according to the road area.At last, a data transportation interface is designed to communicate with cameras. The ADSP-BF561 is the core chip of the data transportation interface. The road detection program sends control commands to the camera and receives images from the camera via this data transportation interface. The architecture of the interface and the implementation of kinds of communication protocols are presented. The uCLinux operation system is installed in the ADSP-BF561. The device driver programming on uCLinux is proposed in this part.