Design of simulation platform of intelligent vehicular visual odometry based on CarSim and Matlab

The visual odometry system based on features mainly consists of the feature detection and tracking module and the pose estimation module. In order to analyze the performance of relative pose estimation algorithms which consider vehicular kinematic constraints, a simulation platform of on-board visual odometry is built using Matlab and CarSim software. This simulation platform, which based on imaging principle of cameras and two view geometry, is composed of vehicle motion simulation module, imaging simulation module, data display and analysis module. The testing object is the relative pose estimation module in the visual odometry system. In this simulation system, the characteristics of vehicular motion are fully considered, and novel thoughts and techniques are put forward to research the application of vehicle kinematic constraints in the visual odometry system. A novel internal pose estimation algorithm is tested on this platform. The experiment results validate the proposed method.