| It is the base and key technology for vision system of tomato-picking robot achieving object-localization and coordinated movement of mechanical arm to identify mature tomatoes in natural growth phase from complex background and to determine the center position and profile size. On the bases of fully understanding international research achievements in the area of fruit picking, the chief purpose of this paper is to study object-localization technology of tomato picking in natural growth phase. The main goal is by imitating human's binocular information through dual CCD camera to perceive object position information in order to locate tomato-picking position, which establishes the base of intelligentizing fruit picking.In this paper, the binocular vision hardware system of the harvest robot was built up. Firstly, the calibrated experiment table was designed, produced and installed, based on location experiment character. Make sure the discrepancy between the camera guide track oriental hole and calibrated board was 5mm, so that the image coordinates would be paralleling moved. Thus, the characteristic point's distribution got improved, and the precision and reliability greatly enhanced. Secondly, the locating mathematic model based on binocular tomato location was brought forward. After the camera calibrating experiment, the matching mark points in the image axis were picked up. The mapping relation between the mark points in the image axis and mark points in the world axis was directly built up, according to the experimental need, in which the calculated inside and outside parameters of the camera was concealed.Thirdly, the mathematic processing method of neuro net was adapted to the mapping relation in the calibrating experiment between mark points in the image axis and mark points in the world axis. By the data processing method of neuro net, the mathematic description of mapping relation was flexibly calculated out. the obtained position model would include every primary and secondly factor inside and outside the camera, so that the subordinate factor would not be neglected, which would result in position precision descendent in mathematic modeling. Finally, the marking light source luminaries needed in building up tomato orientation experimental system was found out in the experiment. The homogenous light source most fit for tomato oriental character was chosen out from seven homogeneous light sources, which were got from physics triple prism refraction. The validity of marking light source... |
