Research On Target Recognition And Location System For Chrysanthemum Picking Robot

As a kind of traditional natural beverage, Hangzhou white chrysanthemum is favored by an increasing number of people in the world and its sales is increasing year by year. In order to decrease harvest labor costs, enhance working efficiency and fulfill the working requirements for automatic Hangzhou white chrysanthemum picking, the design and research of picking robots and its target recognition method for Hangzhou white Chrysanthemum have become an important link of whole process mechanization and intelligentization of Hangzhou white chrysanthemum production and planting in China.On the basis of machine vision technology, with respect to application demand for Hangzhou white chrysanthemum picking robot, in accordance with present conditions of study on fruit and vegetable picking robot, Hangzhou white chrysanthemum identification and orientation system is proposed in this project by application of mechanical system design theory, machine learning, digital image processing and etc. Main content and results of the study is shown as below:1) The preprocessing method towards the image of Hangzhou white chrysanthemum has been studied. The pretreatment method of chrysanthemum images are studied. In the HSV space, the method base on vector median filtering and morphology is used to remove noise, hole and burr.2) Fast fuzzy clustering partition method based on S component is proposed with respect to characteristics of image of Hangzhou white chrysanthemum and on the basis of traditional fuzzy clustering partition algorithm. Fast fuzzy clustering partition method based on S component is used to make traditional fuzzy clustering partition algorithm improved in its time efficiency, reduced in its iterations and clustering center calculations. It averagely takes 0.4s to process one image of 720 pixel×480 pixel and correct recognition rate is able to reach 85%.3) According to the requirements of vision system used for Hangzhou white chrysanthemum harvesting robot, a binocular stereo vision system is designed. Based on theimaging geometric model, the localization principle of the vision system is analyzed. Binocular stereo vision system is calibrated by Zhang Zhengyou's plane calibration method together with APP calibration tool provided by Matlab. It is proved in test that calibration by self-made white and black checkerboard and APP calibration tool kit makes it convenient and accurate to obtain internal and external parameters and deformation coefficients of two cameras in binocular vision system.4) With respect to characteristics of Hangzhou white chrysanthemum zone, characteristics matching method based on barycenter is adopted; barycenter is taken as a characteristic point of characteristics matching; and barycenters of each Hangzhou white chrysanthemum zone is calculated in accordance with results of image segmentation at earlier stage; identical points are matched by characteristic points matching method based on barycenter and epipolar constraint. Then, the fruit positioning was done throughout matching the fruit regions. As the precision test show, when the baseline is 50 mm, and the distance from target to baseline is 150-450 mm, the error between the calculated and measured distance is less than 14 mm, which can meet the requirement of picking accuracy.5) The whole Hangzhou white chrysanthemum harvesting robot is systematically designed, including mechanical arm, picking end executor, computer control system, etc. And according to the requirements and work flow of Hangzhou white chrysanthemum harvesting robot, a control system of Hangzhou white chrysanthemum harvesting robot is developed based on the Windows development environment, development tools of Microsoft Visual Studio 2013, Open Cv and the MFC(Microsoft Foundation Classes) library.To summary, this paper constructed a identification and positioning system for Hangzhou white chrysanthemum picking, observed reasonable methods of image preprocessing, fruit regional segmentation, recognition, extraction and location for Hangzhou white chrysanthemum. All of these can not only provide a reference to the recognition and location of vision system, but also make a foundation for picking robot.