Detection Of Obstacles In Grass By Using Camera And Laser Range Finder

With the development of science and technology, modern robots have been widely used from manufacturing industry to many different fields such as military application, civilian application, service industry and scientific research. They are capable of performing disgusting, risky or monotonous tasks instead of human beings. In the area of robots, camera and laser range finder are two very important sensors and can be found in almost every robot. Color information provided by camera and range information provided by laser range finder are complementary to each other, thus fusion of these two kinds of information has become current research hotspot. This dissertation made use of a joint system of a camera and a laser range finder, and carried out deep study on two key issues of navigation of outdoor robots, which are joint calibration of camera and laser range finder, and detection of obstacles in grass.The dissertation first introduces the concept, theory and method of multi-sensor fusion. Layer model of multi-sensor fusion, and methods used in each layer are emphasized. Methods in joint calibration of multi-sensor are also introduced.Aiming at the system of camera and laser range finder we use, the dissertation studies and raises a new method of joint calibration of this system. Place one plane target in any different positions a few times, and collect information by camera and laser range finder, and calibration of each sensor and joint calibration of multi sensors can be done successively. Different from traditional methods based upon correspondence of key points or edges, this method achieves result by correspondence of distances between sensor's principle point and target plane, and well solves the problem brought along with low accuracy extracting key points or edges.After that, the dissertation studies the detection of obstacles in grass, based on the joint system of a camera and a laser range finder. Grass can't be considered as obstacle when robots move in grass field, so the effect of grass shadowing must be removed to detect the true obstacle. Using max value filter and picking out small absolute value of the second derivative of the range value in information provided by laser range finder, flat obstacles can be extracted in the environment. And scattered misjudged obstacles can be further eliminated by the data of multiple scan planes and multiple scan pictures. At last, the exact place and size of obstacles is fixed by fusing color information provided by camera.