Sports‐related brain injury is a pressing issue, particularly in high‐impact sports like ice hockey where impact velocity plays an important role in determining the magnitude of head impacts and subsequent risk of injury. However, existing methods for measuring impact velocity, such as GPS tracking and manual video analysis, are costly making them inaccessible, especially for youth leagues. This study introduces an automated, cost‐effective method using computer vision to determine player velocity from 2D video. The initial step involves localising the field, achieved through a novel approach employing YOLOv5 to detect specific landmarks on the ice surface. With a dataset of over 9,900 annotated images, YOLOv5 demonstrates exceptional performance, achieving an F1 score and precision‐recall of 0.99 at an 80% confidence level, and mAP scores of 98.5% and 64.5% at IoU thresholds of 0.5 and 0.5:​0.95, respectively. By detecting at least four landmarks per frame, homography matrices were calculated to obtain a top‐down view, completing the localisation process. This approach achieved an average IoU of 0.96, validating its accuracy in field localisation and demonstrating its potential for improving accessibility and cost‐efficiency in measuring impact velocity in ice hockey.