Follow that human

 作者:漆徐剽     |      日期:2019-03-08 07:16:20
By Duncan Graham-Rowe BIG Brother may not only be watching you, he may be following you too. Researchers in California have developed a robot called an Autonomous Observer (AO) that could lead to surveillance systems that follow people wherever they go. Other systems can’t follow moving objects that are trying to evade detection. If an object disappears from view, the tracking system loses it. To overcome this, computer scientist Jean-Claude Latombe at Stanford University has developed small mobile robots that not only watch targets but also work out their potential escape routes. The robots can then position themselves for an optimum view. This task may be simple for humans but it is a huge challenge for robots since they must “understand” exactly what is around them. To work out where objects are, the AO uses a laser rangefinder to add depth to images captured by a video camera. That way, it constructs a three-dimensional representation of its environment that it can refer to as it follows a target. As an AO tracks a target, it uses this model to look for the structure that the target is most likely to hide behind, such as a wall or a corner. The AO then moves to give its video camera the widest possible field of view, making it harder for any target to escape. And since the AO updates its 3D map four times a second, in tests human and robot targets have found it very hard to escape detection. At present, the AOs are confined to the lab. To navigate, each robot uses a video camera to monitor its position relative to a grid pattern on the ceiling. But, says Latombe, it should be simple to fit each robot with a global positioning system so that it can work outside. AOs could soon be used as supervisors in factories, overseeing other robots and reporting breakdowns. But they could do more. Latombe is now receiving funding from the US Defense Advanced Research Projects Agency to explore the military potential of his tracking system. “Robots could one day be used to scout buildings where enemies might be,” he says. And companies could use Latombe’s tracking software to improve their closed-circuit security systems. “If you have too many cameras, the problem is knowing which one to look at,” Latombe says. With his software in control, however, a security system might be able track intruders as they move through a building,