3D vehicle sensor based on monocular vision

摘要: Determining the position of other vehicles on road is a key information to help driver assistance systems increase driver's safety. Accordingly, work presented in this paper addresses problem detecting front our own one and estimating their 3D by using single monochrome camera. Rather than predefined high level image features as symmetry, shadow search, etc., proposal for vehicle detection based learning process that determines, from training set, which are best distinguish non-vehicles. To compute with camera point consists knowing where horizon projects onto image. However, can change every frame difficult determine. In we study coupling between perceived actual width order reduce uncertainty estimated derived an unknown horizon.