Multi-Classifier Based LIDAR and Camera Fusion

Abstract

We present a sensor fusion system using lidar and camera. We separate the system into two part which is hypothesis generation part and hypothesis verification part. These parts use different single sensors. Hypothesis generation is done using the lidar and hypothesis verification is done using the camera image. In hypothesis generation, we cluster the lidar data and do a perspective mapping to generate the candidate. In hypothesis verification, we used 5-SVMs classifier. Based on the candidate position, the candidate is putted in different SVM. In the result, we compared the result between 5-SVM hypothesis verification and single SVM hypothesis verification. The result showed 2.2% improvement.