3D gesture estimation method and head-mounted display device

The invention relates to the technical field of three-dimensional reconstruction, provides a 3D gesture estimation method and a head-mounted display device, and designs a circular region segmentation method by using a wrist joint point and a middle finger root joint point, thereby accurately segmenting a hand point cloud from a depth image, reducing the influence of the point cloud of the wrist part on gesture estimation, and improving the accuracy of gesture estimation. Meanwhile, in the model optimization matching process, some 2D constraint terms are added through matching between projection points from articulation points in the hand model to the RGB image and articulation points detected in the RGB image and matching between projection areas from the hand model to the RGB image and hand areas segmented by the RGB image, and therefore the 3D gesture estimation accuracy is improved. Under the condition that the finger is shielded, the problem of depth data loss of the finger part caused by inherent defects