Recognition And Match Of Overlapped Oranges From Nature Scene Based On Stereo Vision

With the rapid development of agricultural production, the cost of agriculture labor force will become more and more costly. In recent years, the agricultural application of robot technique have already become popular issue, because of the shortage of the agriculture labor force both in developed countries and developing countries. Different from industrial robot, which works in particular environment, the agriculture robot mainly works in the natural environment, and the agriculture robot has to face more complicated and uncertain circumstance, thus there are more problems to be resolved.Different mature period of the fruits leads to long harvesting circle, thereby the share of manpower fee in the fruit price is very high. For this reason, fruit picking robots are studied by many research institutes recently. However the complexity of nature environment, there are a large number of problems dealing with the location and the mature degree judgment of fruit with robot vision, there are also open issues at the manipulator's picking operation. Hence there is still a long way from laboratory to practical application. Orange is typical spherical fruit with distinct color characteristic. In this paper, we take it as the research example. Binocular stereoscopic visual system has used to match fruits and identify mature fruits. Among them, the key problems we try to resolve involve identification of mature fruits in variational brightness,portion of the fruits conglutinated and overlapped or sheltered circumstance. This is also important to exploit the practical agriculture harvest robot in the future.The researches are summarized as following:1. RecognitionRecognition is in order to discriminate mature oranges from nature scene and prepare for three-dimensional location. According to the analysis of the different condition's images that got from complex backgrounds, and through the compare experiment we find that the Ostu algorithm of the 2R-G-B can recognize mature fruits easily and accurately. Turn the object color image after segmentation to black and white image, through the morphologic operation eliminate the noise, then label the different area, according to the area and the ratio of the least external rectangle's length and width set the threshold, wipe off the other area, then fill the holes and detect the edge. At last, using the optimized Circular Hough transformation (CHT) algorithm simulates the object image centroid coordinates and radius realistically. Experimental results show that the recognition correctness rate is high up to 95%.2. MatchThe purpose of Match is to make certain the 3-D position of the mature oranges. The robot could not pick a mature orange until the spatial position of the orange was got. In order to get the 3-D position of the orange, the binocular stereovision technology was introduced. Triangle algorithm is adopted to locate the fruit in the world coordinate system. Combining with epipolar restriction and sequence restriction, under the two conditions of via the barycenter and simulate the centre of a circle, based on the algorithm of the area,average gray and the gray standard deviation mutual restriction, a new feature match technique was used to find corresponding targets in two images. The algorithm is implemented to sole match. The experiment result demonstrates its feasibility and the match correctness rate is high up to 90%. At the same time, the camera was calibrated. In the end, with the laser range finder emendation, we find that when the work distance is less than 1.3m, almost all the errors are less than±4mm.Our research has made great progress in identification and match of mature oranges and fruit location. The experimental system can accurately identify and locate mature oranges in greenhouse or in open-air, where lightness changes greatly, even though there is partially overlapped or sheltered of the fruits. The researches presented in this paper benefit further practical application of fruit picking robot on vision part. It is also meaningful to improve the international competition in our agriculture field.