Multi-layer segmentation solution to filter the noise of optical flow vectors to assist robots in object recognition inside buildings

In this paper, a qualitative optical flow processing solution is proposed to filter out the noisy vectors. The qualitative processing solution is performed based on the results of the multi-layered segmentations of the optical flow vectors before calculating a filter threshold related to the mean/average of the vectors in the segmented optical flow image area (called quarter). The advantage of the proposed solution is that the filter threshold is calculated using simple arithmetic operations and its value varies according to different noise quarters. Experimental results in an indoor environment will demonstrate that the ability to recognize objects based on computer vision after filtering optical flow noise helps the robot move safely to a desired location.