Object Velocity Estimation Based on Asynchronous Data from a Dual-Line Sensor System

This paper presents a 2 times 256 dual-line vision system for object velocity estimation based on the asynchronous output data. For real-time velocity estimation, a processing method has been developed and implemented on the digital signal processor. It exploits the sparse asynchronous data representation from the dual-line sensor with high temporal resolution (better than 100 mus) and low latency. The processing concept includes the object contours extraction, velocity estimation and scaling. Three approaches are used and evaluated for the velocity estimation. The first and second approaches use the mean and median object detection time, respectively for the object velocity calculation. In the third approach, the statistical linear fit RANSAC is used to extract the object time. This processing concept has been evaluated on object velocities that range from 1 to 22 m/s, for the three velocity estimation approaches and a comparative study has been provided. The experimental results included show a velocity estimation error < 1% using RANSAC.