Hand posture estimation method and device, computer device and storage medium

Abstract

The invention discloses a hand posture estimation method and device, a computer device and a storage medium. The method comprises the following steps: performing depth difference calculation on a depth image to obtain a foreground mask image; processing the foreground mask image to obtain an input image; enabling the trunk branch to perform convolution layer processing and pooling processing on the input image to obtain a main feature map; enabling the classification branches to judge whether the main feature map is a right-hand image or a left-hand image according to features of the main feature map; under the condition that the classification branch judges that the main feature map is a left-hand image, inputting the main feature map into a left hand branch; under the condition that theclassification branch judges that the main feature map is a right-hand image, inputting the main feature map into the right hand branch; calculating the normalized three-dimensional coordinates through the left hand branch or the right hand branch, and outputting the three-dimensional coordinates of the hand joint point through the coordinate conversion formula; according to the invention, the problem of left-right hand confusion or joint point prediction failure is solved, and the recognition accuracy of the normalized three-dimensional coordinates of the hand is improved no matter for the left hand or the right hand.

Description

Technical field [0001] This application relates to the field of computer vision technology, in particular to methods, equipment, computer equipment and storage media for hand posture estimation. Background technique [0002] With the development of human-computer interaction technology, human-computer interaction methods such as keyboard, mouse, touch screen, etc. have been difficult to meet the requirements of users in many emerging fields. In augmented reality (Augmented Reality, abbreviated as AR), virtual reality (Virtual Reality, abbreviated as In scenarios such as VR) and remote control, users are more inclined to use wireless, non-contact ways to achieve human-computer interaction. On the other hand, gesture interaction technology based on computer vision technology allows users to get rid of complex interactive devices and use specific gesture actions to issue instructions to the machine, which is convenient and quick. Therefore, the development of computer vision technol...

AI Technical Summary

Problems solved by technology

[0003] The core of gesture interaction technology is to estimate the three-dimensional coordinates of each joint point of the hand in space, judge the user's gesture through the three-dimensional coordinates of the joint point, and complete dynamic gesture interactions such as air click and air slide by obtaining the position of each finger. The fully connected layer is usually used to estimate the hand pose. Since the joint points of the left hand and the right hand are distributed in different positions in the input image, and the fully connected layer is difficult to distinguish the left hand from the right hand, it is easy to cause confusion between the left and right hands or the failure of joint point prediction.

[0004] In the related technology, it is difficult to distinguish the left hand from the right hand in the fully connected layer, which may easily lead to the confusion of left and right hands or the failure of joint point prediction. At present, no effective solution has been proposed.

Images