Research Of Fast Target-Detection Technology Used On Robot Vision Based On Biology Vision Principle

Robot vision is one of the most important apperceive modes of robot intelligence. Although the research of robot vision has acquired much fruit, the ability of existing robot vision system is not equal to identify and understand in complex background surrounding request high real time. The auto target-detection takes complex image surrounding as object is one of the difficult problems in it, which requests not only the exact and credible target-detection and track, but also real-time and fast recognition and reaction. Therefore, there is important academic significance and applied value in the research of method of vision information processing and target recognition and tracking based on biology vision principle.The main content of research and result includes:1. The correlative fruit of neurophysiology and psychology in the principle of human vision processing is analyzed and summarized. Introduce the human vision gateway, the principle of ganglion cell recept field, the principle of vision centrum, principle of color apperceive and the principle of hazily storage and recognize of human vision to color and shape character.2. According to the inspirer inspired by biology and the principle of multi-channel processing of biology vision, a method that process color and edge and movement information parallelly by multithreading is researched. Aim at the difficult problem that the threshold of traditionary robot vision edge-distill is influenced easily by the change of illumination, bring forward On recept field edge-distill template, which not only increase the speed of edge-distill but also adopt zero as threshold in order to depress the difficulty in threshold choice. Based on fuzzy theory, research of the method of color determinant, which gives the ability adapt to different illumination condition to robot vision system and has much applied value to robot vision to achieve color constant. Analyse the method of mobile target detection, add the movement character to target-detection to increase the veracity of detection.3. Take robot vision system target-detection as object, this paper analyses the demarcate method of robot vision system. Improve the quality of the binary image which is the result of the color determinant by morphologic method. Analyse the normal template matching method and research of the application of this method in shape determinant. Then, according to the result of experiment, chose the morphologic processing method with better effect, which has also scale and orient invariability, into shape determinant.4. Design and construct the target-detection and tracking experiment system based on vision, then excogitate multithreading vision information processing software by graphic language Lab VIEW in order to validate the theory and method bring forward above. The result of experiment indicates that the method in this paper achieve very high speed whether it is used to detect and track to object with random shape.Mass result indicate that this method is able to detect truly and fleetly the position of the target with any shape. The detection speed is 6-7 frames per-second to the image with 640X480 pels and the precision is 0.3mm, which satisfy to the demand of navigation of mobile robot and manipulator. Moreover, this paper proves that it is an effective approach that uses the fruit in the domain of neurophysiology and psychology for reference to resolve the difficult problem in robot vision.