Design And Research On The Vision Based Autonomous Instrument Recognition System Of Surveillance Robots

As a kind of non-contact sensor, image sensor with high precision is widely utilized in security, medical, industry and many other fields. Power substation surveillance robot is comprise of autonomous navigation and integrated with visual camera, infrared and sound sensors. This kind of robot has been increasingly used in urban security inspection, intelligent vilage security inspection and checking in power station. For the situation of high voltage and the corresponding high risk in the power substation, it is necessary to use security surveillance robot instead of manual inspection. Considering about the diversity of instruments in substation, the high-precion visual algorithm must be introduced into image processing. This is the only way to manage the power substation in unmanned and intelligent method.According to the above mentioned problem, this thesis adopts the image preprocessing algorithm which based on visual significance analysis into the reducing of template traversal area, and combined this method with cloud robotics technology to improve the processing speed and accuracy.With purpose of catch the postion of instrument, the image transformation of the absolute dual quadric based camera self-calibration has been utilized. This method is an effective way to pretend the offset of the object, and then the PTZ system can be controlled with high-precision. The advantage of this method is the simple structure, reliable precision, without the stereo matching.Instrument reflective situation will seriously affect the robot reading accuracy. In this thesis, the visual attention based bright spot detection has been adopted on the instrument image processing, and then the flare have been removed at the same time. Besides, this thesis recognate the number of instrument by estimate the angle between pointer and some on fixed feature, whose contours are all acquired by Fourier descriptor for its low-sensitivity to light. As for the perspective distortion, this method has also realized inhibiton in some degree.Additionally an experiment platform has been built. A two-degree-freedom mechanism have been designed with the corresponding hardware platform. The user operation interface is written in Visual Studio 2010. The error of the measuring system is analyzed in theory. Experiments are carried out on the rotating platform.