Research On Peeled Mandarin Segments Sorting Device Based On Machine Vision

During the production of canned mandarin, some mandarin segments have defects like membrane residues, seed residues or segment damage that affecting products quality. Currently factories use manual inspection to reject defective segments. Using manual inspection may cause issues like difficult to find labors, high costs, hygiene problems etc. An automatic sorting device for peeled mandarin segments based on machine vision was developed, and related inspect algorithm was studied in this thesis.A machine vision inspect system was designed for these three kinds of segments defects. Several cameras, lenses and lights have been studied for their imaging performance. We discovered that ring light can produce more uniform illumination when flat light was better for seed inspection. An inspect platform was built with white ring light and red flat light and a 658x492 pixels color camera. Images acquired were acceptable with clear defect characters.Based on LabVIEW and Vision Development Module, machine vision algorithm as graying, binarization, image enhancement and defect segmentation etc were studied and compared. An inspect and control program was written and tested on the inspect platform. Result shows the accuracy of inspection program reached 99.1±0.5% and the average time consumption is 17±5ms. The inspect speed is able to achieve 64±19 pieces per second.A mandarin segment sorting device has been designed. To resolve the problem of segments overlapping, four kinds of segments separation devices were designed and tested. The best method can reduce the overlapping rate to 0.27±0.18%. Two kinds of conveying device and an eliminating device were studied for the fragile segments, which can decrease the breaking rate of segments to 0.18+0.15%.The mandarin segments sorting device model was built to test the performance of vision algorithm and machines. The results indicated that the max speed of the automatic sorting device was 30 pieces per minute. The identification accuracy reached 99.1±0.3% and the breakage rate of mandarin segments was 0.11±0.05%.