Study Of The Visual Servoing System On Intelligent Robot

As an advanced product of optical, mechanical and electronic integration, sensation and intelligence are the first and key problem of the robot. Vision is the most important human sense. Vision sensor has many features, such as obtaining information largely and fast, high precision, tending to integrate the design information with control information, and feeling the environment non-contact. Therefore, robot visual servo system has an important status in robotic research and application, also has a decisive influence on robot intelligence.In this article,we mainly studied the current problems exiting in the robot visual servo system, and do further research in camera calibration, image edge detection and recognition, design of visual controller. Here, lists the innovations and achievements.1. With the neural network's stability of study, adaptive and approximation of nonlinear function, the mapping relationship between spatial point and image point is established, and the neural network-based camera calibration is completed. Compared with traditional calibration method, the neural network-based method highly improves the precision and robustness of calibration.2. Considering the complex work situation of most robot visual servo system, combine the information fusion and fuzzy edge detection, put forward a fuzzy edge detection algorithm that is efficient in high noise image process. Traditional fuzzy edge detection algorithm is purely based on image enhancement. While fuzzy edge detection algorithm makes use of multi-information of pixel point as edge detection information, and fuzzy logic to combine these information, eventually makes the edge information more prefect and valid. Experiment shows the obvious advantage of the fuzzy edge detection algorithm in high noise image edge detection.3. Taking the advantage of the rapidity of optical wavelet transformation and anti-noise stability of Wavelet Multi-scale Product, combine both together to enhance the real-time property and robustness of image process in visual servo system.4. The image pattern recognition is realized by invariant moment and BP/ART-2 neural network. Compared with traditional pattern recognition method, the neural network-based method has strong fault tolerance and adaptive ability. Furthermore it adopts parallel working style, and has fast recognition speed.5. As we know, current visual servo systems only converge in local area. According to the hybrid dynamic system theory, A switched visual servo system is established using homography-based and affine approximation controller. Simulation results show the validness of the switched system, and a new idea in the design of controller is gived.