Automatic Body Measurement for Straight-Tube Mountain Bike Size Selection Using Closest Size Classification

The lack of bike expert personnel in bike shops challenges new mountain bike riders to choose the correct bike size. Moreover, there are currently no automated methods specifically designed for mountain bike size selection. This paper proposes a method for selecting the appropriate mountain bike size based on top tube length and seat tube length. This study focuses on mountain bikes with straight-tube frame bikes. The process involves three steps: human segmentation, body size measurement, and bike size selection. U-Net with batch normalization layers and morphological operations is used to accurately segment the human area, followed by measurements of the rider’s arm length, inseam height, and torso length. Finally, a bike size is selected using a closest size classification. Five experiments validated the method: 1) human segmentation results, 2) feature point detection results, 3) body size measurement results, 4) top tube length and seat tube length measurement results, and 5) bike size selection results. The human segmentation achieved a 99.24% F1-score on the test dataset. The average relative absolute errors of the inseam height, arm length, and torso length are 1.93%, 2.37% and 3.12%, respectively. The average relative absolute errors of the top tube length and seat tube length are 2.00% and 1.93%, respectively. The accuracy rate of bike size selection for all three models is 90.80%. The novelty of this work lies in its end-to-end automation, eliminating the need for expert assistance or physical markers. This makes it a pioneering contribution to the field of automated bike fitting, with significant implications for consumer convenience and retail efficiency.