Development of a probabilistic perception system for camera-lidar sensor fusion

Abstract

Multi-modal depth estimation is one of the key challenges for endowing autonomous machines with robust robotic perception capabilities. There has been an outstan- ding advance in the development of uni-modal depth estimation techniques based on either monocular cameras, because of their rich resolution or LiDAR sensors due to the precise geometric data they provide. However, each of them suffers from so- me inherent drawbacks like high sensitivity to changes in illumination conditions in the case of cameras and limited resolution for the LiDARs. Sensor fusion can be used to combine the merits and compensate the downsides of these two kinds of sensors. Nevertheless, current fusion methods work at a high level. They processes sensor data streams independently and combine the high level estimates obtained for each sensor. In this thesis, I tackle the problem at a low level, fusing the raw sensor streams, thus obtaining depth estimates which are both dense and precise, and can be used as a unified multi-modal data source for higher level estimation problems. This work proposes a Conditional Random Field (CRF) model with multiple geo- metry and appearance potentials that seamlessly represents the problem of estima- ting dense depth maps from camera and LiDAR data. The model can be optimized efficiently using the Conjugate Gradient Squared (CGS) algorithm. The proposed method was evaluated and compared with the state-of-the-art using the commonly used KITTI benchmark dataset. In addition, the model is qualitatively evaluated using data acquired by the author of this work.