Computer Vision-Based Automated Archery Performance Logging System

In the sport of archery, a stable shooting stance is crucial to obtain high-precision shooting. However, wearable motion sensors might interfere with the shooting performance of an archer. In this paper, a markerless-based motion capture system using a deep learning approach is proposed to retrieve the keypoint of archer shooting in a continuous shooting phase. The correlation of shooting posture is studied with the scoring precision and grouping. Studies are conducted on two types of archery shooting sessions using recurve and compound bow. The joint angle profile can be decomposed into six phases, i.e. stance, pre-draw, draw & anchoring, transfer, aiming, and follow through. Important parameters and joints, particularly the angles and their consistencies, are analyzed to study archers shooting performances. An algorithm is proposed in this study to automate the determination of the archer’s anchoring and release sequence. Several qualitative results of joint kinematics and parameters and visualization are presented to further understand the bio-mechanical aspect of archery. In addition, this study reveals a direct positive correlation in posture consistency between arrow precision and grouping. Hence, it indicates a consistent shooting posture is a significant attribute of an archer’s performance.