Research On Real-time Preprocessing Method Of TOF Low Resolution Point Cloud Data For Human-computer Interaction

With the development of human-computer interaction technology,spatial information based human-computer interaction technology is in hotspot.Compared with multi-view cameras and structured light based depth sensors,active light source solid-state lidars have advantages,such as small size,high refresh rate and light tolerance.This paper mainly focuses on the data preprocessing of low-resolution TOF multi-point sensor and its application in human-computer interaction.For the noisy TOF depth image,traditional filtering algorithm leads to edge detail distortion.We proposed a bilateral filtering preprocessing method for low resolution of TOF measurements.Then,a fast interpolation preprocessing method based on edge gradient was proposed.For gesture interaction with low-resolution TOF depth data,we performed gesture segmentation and fingertip detection test,preliminary results of gesture interaction was presented.We will use low-resolution solid-state TOF sensors to realize the hand gesture recognition.