Research Of The Vision System For Apple Harvesting Robot Woking In Natural Environment

It is of great significance to develop the mechanized harvesting technology andresearch the fruit picking robot. Only if the vision system of picking robot was improvedperfectly, can the robot pick off the fruit precisely. However, due to the uncertainty of lightconditions in operating environment and the complexity of background, there is still not apractical and stable vision system of picking robot.In this paper, red apple which is grown in far-ranging area is the main object of study.Study the essence of why the vision system is susceptible to interference. Reveal the effectlaw of light when taking the picking images. Structure the path of processing the imagewhich is aimed at natural environment. Optimize the identification and positioning systemby design and implementation systematic experiments. The main contents and methods areas follows:This paper research the fruit images took in natural light and proposes a newsegmentation method for fruit images which combines the dynamic threshold and fixedthreshold: firstly segment the grey fruit image in R channel by the thresholding calculatedby iteration, then segment the colored subgoal image by fixed thresholding, finally binarysubgoal images could be obtained. Validate and optimize the image segmentation andtarget recognition algorithm by recognition experiment.After studying the geometric features of apple, circular form is selected to fit theprofilogram of this fruit and the centre of the circle is thought as the centre of apple, andthe direct basis for locating fruits is obtained.In order to determine the spatial position of fruit, this paper used binocular stereovision system. Match the apples from the different images by the principle of right-leftorder consistency and uniqueness, and get the image coordinates of the same apple.Calibrate the binocular camera and get the system parameters, then get these parametersinto the space coordinates formula to determine the center coordinate of every fruit. Getthe best operating parameters for the binocular camera system used in this paper bypositioning experiments.Through systematic and in-depth research, this paper constructs a widely used visual recognition system. This system can wok in natural light. This paper provides a theoreticalbasis for significantly improving the accuracy and stability of picking robot vision system,and the technical support would made the mechanization of harvesting coming into realizemore quickly.