Multi-camera collision detection between known and unknown objects

摘要: Today, real-time collision detection is a basic demand for many applications. While tests between known (modeled) objects have been around quite while, of with dynamic, unknown (sensor-detected) remains challenging field research, especially when it comes to requirements. The test described in this paper based on several stationary, calibrated video cameras, each supervising the entire 3-dimensional space shared by and (e.g. humans robots). Based their images, potential collisions any (future) configurations priori dynamic obstacles are detected. Occlusions caused (such as robot or machinery set-up within workspace) detected addressed safe manner exploiting geometrical information epipolar line geometry cameras decision fusion process. algorithm can be parameterized adapt different application demands. Experimental validation shows that behaviour possible presence highly they occur robots share same workspace accomplishment task. In effect, vision-based safely used human-robot cooperation, intrusion detection, velocity damping, obstacle avoidance.