Feature Matching Algorithm Design and Verification in Rotates Camera Normal Region Based on ROS System

Intelligent vehicle is one kind of popular research field at present, and the environment perception of unmanned vehicle, especially the detection and tracking of target objects, is indispensable for it's development. In this paper, we built an image acquisition simulation environment based on the robot operating system (ROS) and Gazebo, and designed a algorithm to realize the feature matching in normal region of the rotating camera. Different from the common matching algorithm, this paper places the target object at the imaging center of the right camera through parameter setting and rotates the left camera in the normal ROI region to obtain the images of the two cameras for feature matching. In order to verify the accuracy of the algorithm, during the rotation of left camera this paper selected the symmetrical angle with the camera on the right side, as well as 10 angles for the change of the fixed step size on the left and right sides of the symmetrical angle to acquire images, and analyzed and compared the angles and matching rate as parameters to verify the accuracy and stability of the matching algorithm.