We developed a green soybean sorting machine prototype to improve the efficiency of green soybean processing. Image processing and transparency images were used to identify damage to green soybean pods, and investigate its sorting accuracies. The green soybean sorting machine prototype was 1700mm long, 500mm wide, and 1500mm high, comprising two conveyer belts, an arranging and transportation mechanism, finder units, processing units, and a separation system. The arranging and transportation mechanism, which comprises an arranging roller, some arranging rolls and transportation guides, produced the most stable green soybean transport posture among all types. The green soybean sorting machine prototype, which included that mechanism, showed a five-fold higher sorting rate than manual sorting. Finder units comprising image processing units and fiber sensor units detected seed maturity and green soybean damage that produced color change points in the pods. When putting the pods on a conveyer belt one-by-one at a laboratory, sorting accuracy (η) indexes of the green soybean sorting machines prototype were, respectively, 0.51 for the finder unit, and 0.80 for the fiber-sensor accuracy (η
s), and 0.66 for the image processing unit accuracy (η
i). Sorting tests at a production center were 0.11 at the average η because of effects on degradation for the transport posture and smaller damaged areas of pods. Nevertheless, this green soybean sorting machine prototype was judged as having 0.75 sorting accuracy for good pods (η
a), which was similar to manual sorting : 63-81% judged as good pods (
WgPg).
View full abstract