Improving Vehicle Detection in Point Cloud Data with Novel Features

To understand one of the autonomous tasks for self-driving car is to classify the vehicles. This paper presents a pipeline of Lidar point data's clustering and classifying, basically conducted by combining the grid-occupy segmentation and Support Vector Machine learning method. To improve the classification, a novel feature related to vehicle's plate intensity characteristic is investigated. After point cloud clustering, different classifiers were utilized to evaluate the accuracy and our proposed method produced better performance.