Stereo Vision Based Worker Detection System for Excavator

It always has been a serious issue to improve safety in mine sites, demolition work sites, and civil engineering work sites. Especially, collision accidents caused by construction equipment damage not only the productivity of operations but also the health and life of workers in those work sites.

To solve this problem, a variety of worker auto-detection means are developed so far, but they have not still become widely used in actual worksites. There have been two controversial points in the prior methods. One is that they were not enough in respect of the detection accuracy and certainty, and the other one is that the information of the detection results was not comprehensible to the operator of equipment.

In this paper, the functional requirements of the obstacle detection system in earth moving work site is redefined at first, and then, it is proposed to apply a stereo vision to the obstacle detection system for an excavator.

The system consists of horizontally arrayed two digital cameras with low distortion lens, an image processing controller, and a monitor for an operator in the cab. Stereo vision calculates the depth image from images of two cameras and detects the presence of obstacles, its distance, direction, and dimensions. Sensed obstacles are informed to the operator in the monitor image.

Preliminary experiments are conducted with a stereo vision system mounted on a real equipment in a demonstration site. The result shows that two controversial points of the conventional technologies are improved, and that the applicability of this system is validated.