Golf Ball Detection and Tracking Based on Convolutional Neural Networks

Abstract

With the rapid growth in artificial intelligence (AI), AI technologies have completely changed our lives. Especially in the sports field, AI starts to play the role in auxiliary training, data management, and systems that analyze training performance for athletes. Golf is one of the most popular sports in the world, which frequently utilize video analysis during training. Video analysis falls into the computer vision category. Computer vision is the field that benefited most during the AI revolution, especially the emerging of deep learning. This thesis focuses on the problem of real-time detection and tracking of a golf ball from video sequences. We introduce an efficient and effective solution by integrating object detection and a discrete Kalman model. For ball detection, five classical convolutional neural network based detection models are implemented, including Faster R-CNN, SSD, RefineDet, YOLOv3, and its lite version, YOLOv3 tiny. At the tracking stage, a discrete Kalman filter is employed to predict the location of the golf ball based on its previous observations. As a trade-off between the detection accuracy and detection time, we took advantage of image patches rather than the entire images for detection. In order to train the detection models and test the tracking algorithm, we collect and annotate a collection of golf ball dataset. Extensive experimental results are performed to demonstrate the effectiveness of the proposed technique and compare the performance of different neural network models.