A palm tracking method and system based on depth images

Abstract

The invention discloses a depth-image-based palm tracking method and system. The method comprises: a first depth image among a plurality of depth images containing gestures is transformed into a 3D point cloud and an estimated palm point is obtained; regional growth is carried out by starting with the estimated palm point in a geodesic line distance threshold value range and a palm connection set is obtained; a boundary point set of the palm connection set is obtained; a geodesic line distance between each point in the palm connection set and an inverse regional growth seed point beyond the palm connection set range is determined; a tracking point is determined based on the geodesic line distances and the boundary point set; and a palm is tracked based on the tracking point. Therefore, real-time and robust palm tracking is realized; and a palm tracking failure can be discovered timely and can be processed timely.

Description

technical field [0001] The present application relates to the field of image processing, in particular to a method and system for palm tracking based on depth images. Background technique [0002] Traditional RGB-based gesture tracking is greatly affected by background environment, lighting and other factors. The depth map can obtain the position information of objects in 3D space, so it has great advantages in target tracking. Traditional gesture tracking methods based on depth maps include detection-based tracking, connected-set-based segmentation and tracking, and so on. However, because the hand shape changes a lot during gesture tracking, the speed of the hand is fast and slow, and it is interfered by the background, body, head, etc., these methods can achieve gesture tracking under certain conditions, but the robustness is poor, that is, the continuous tracking ability poor. For example, a method is disclosed in the reference (Chia-Ping Chen, Yu-Ting Chen, Ping-Han L...

AI Technical Summary

Problems solved by technology

The growth point of this method is unstable, and the hand area often grows incompletely, and the growth is easily affected by objects such as the body near the palm. When the movement speed is too fast or it is connected to other areas, it will cause the target tracking to fail and cannot be recovered.

Images