Accurate delineation and characterization of tree branch structures are vital for assessing the status of trees, forecasting the risks of diseases, and formulating conservation strategies. Current segmentation methods struggle to effectively handle the intricacies and diversity present in branch structures, particularly branches that are slender or obscured by dense leaves. To address these issues, this paper introduces the Occlusion-perception and Dense-dilation Network (ODNet), an innovative network designed to enhance tree branch segmentation and reduce occlusion impacts. Firstly, the Dense Dilated Module within ODNet enhances adaptability to diverse tree scales, addressing species variations and complex branch structures. Secondly, drawing insights from tree growth patterns that usually grow upwards and outwards, we designed Orthogonal Perception Module (OPM) . OPM offers a comprehensive perspective of tree branch structures, which can mitigate the prediction discontinuities caused by leaves occlusion. The quantitative analysis and visual comparisons indicate that ODNet achieves better results than existing methods.

A ReID-based multi-object tracking (MOT) network with adjacent frame enhancement is proposed, which is constructed with center-based detection network. To enrich the ReID features, our method treats the adjacent frame as data augmentation of the current frame, implicitly uses the network to achieve data alignment for each object at different frames, and then uses the aligned adjacent features to enhance the ReID. In the training phase, the method adds a location prediction branch to guide the network to learn the feature correspondence of the same object in two frames. In the inference phase, our method uses historical tracking to construct a heatmap of interest of the adjacent frame to inform the network of the locations of all objects in adjacent frames, then the network autonomously mines the features of each object in these two frames. Our approach performs better in MOT challenges than the existing multi-object tracking network.

Immersive Virtual Reality (VR) systems that combine Head Mounted Displays (HMDs) and a position tracking system support the multiple users or participants to collaborate in the same physical space for a large-scale virtual environment. Because the multiple users sharing the same physical space are in a dynamic state, the key technique of multi-user VR system is how to solve the problem of potential collisions among the users who are moving both virtually and physically. In order to better solve the collision problem caused by such dynamic changes, this work presents a new strategy of multi-user redirected walking using dynamic artificial potential fields, which generates repulsion to ‘push’ users away from obstacles and other users, and uses gravity to ‘attract’ users to an open or unobstructed space. In this method, the users not only get repulsive forces from walls, but also from other users and their future states that are called avatars. At the same time, the users will get gravitational force from the steering target. The target selection considers the size of open space, the distance between the steering target and the boundary of physical space, and the distance between the steering target and the center of the physical space. Therefore, the system can steer users to an open area in the physical space to further reduce collisions. To verify the validity of our method, we developed a software to statistically analyze the influence of different factors, such as the physical space size and the number of users. Data from experiments shows that our method reduces the potential user resets by about 20%.