Human-body trunk identification device and method, and terminal-point image detection method and device

Abstract

The invention proposes a human-body trunk identification device which includes a target extraction module used for obtaining a body-shape contour of a line dashing image and determining a minimum external-connection rectangle of the body-shape contour so as to obtain a target area in the line dashing image; a body-shape segmentation module used for segmenting a body shape from the target area; and an identification module used for identifying a trunk part from the body shape. The device accurately positions the target area and effectively eliminates interferences of lower limbs, arms and a head and the like through body shape segmentation and thus a human body trunk part of the line dashing image can be identified rapidly and accurately. The invention also proposes a human-body trunk identification method, and a finish-line detection device and system.

Description

technical field [0001] The invention relates to the technical field of competitive game detection, in particular to a human body torso recognition device and method, and an end point image detection method, device and system. Background technique [0002] The finish line image detection system is now widely used in various modern track events. The system can capture the linear image sequence of the track runners when they cross the finish line. Since the line scan images are taken at a fixed point at a specific moment, these line scan image sequences need to be stitched together in sequence to obtain the players' line-crossing moments. Manual interpretation of the finish line images requires professional referees to identify and mark the exact line-crossing position of the players on the line-array mosaic images. However, manual judgment has low real-time performance and lack of objectivity, so it is difficult to meet the requirements of modern high-level competition. [0...

AI Technical Summary

Problems solved by technology

Mikolajczyk et al. proposed a method for detecting limbs, but this method only judges whether the head exists

Yoon proposed a robust human head detection method for target tracking, but this method is based on the assumption that the contours of the head and body present an Ω shape, and the universality is poor.

Images