Implementation of optical flow measurement system with an embedded processor

Optical flow estimation is one of the measuring method of object motion. It is important to prove that optical flow measurement system can perform in small scale computer to use as visual sensor. In this research, processing speed of optical flow measurement system with a single board computer Raspberry Pi was evaluated. Calclation time of optical flow estimation program based on spatio-temporal differentiation method with eigenvalue decomposition was compared with those using built-in optical flow function in OpenCV library (Lucas-Kanade method and Horn-Schunck method). As a result, the program was about the same processing speed as HS method, and took about six times as long as LK method. It is also shown when an object moves at a velocity of about 10 pixels per frame or more, output results apt to show wrong velocity vectors. Processing speed is to be improved by selecting optimum pixels required for velocity estimation. It will be necessary to compensate for velocity so that it is able to estimate optical flow at a high speed.