An implementation of distance measure for dynamic scene

Abstract

To avoid calibration complex steps and image matching inaccurate, an implementation of distance measure for dynamic scene is proposed. Two cameras are fixed at the lens mount and responsible for getting real time images, first, detect two-dimensional coordinates of target object in the images which got by detection algorithms. According to the camera perspective projection model, there is a mapping relation between two-dimensional coordinates of target object in the image and three-dimensional coordinates of mapping point in the calibration board. Then, calculate three-dimensional coordinates of mapping points in calibration board based on calibration board height and width, distance between two cameras and distance between camera and corresponding calibration board. Later, connect camera and corresponding calibration board to make up two space lines and they will intersect in the target object position. Last, calculate distance between target object and the camera. Experiments show good accuracy and high real-time performance.