Abstract
It is desirable to develop the visual sensor which can detect fruit for agricultural robot and plant factory which have been studied recently. In this paper, fruit discrimination and recognition were experimented in order to develop cucumber fruit detecting visual sensor.
Generally, cucumber fruit has so similar color to its leaf and stem that it is difficult to discriminate them by use of R, G, B signals of color T.V. camera. In this experiment, image was input by using a monochrome T.V. camera and two interference filters which had sensitivity not only in visible region but also in infrared region for discrimation. Secondly, binary image was made after calculation between the two gray level images. Thirdly, fruit was recognized by features of the binary image and line image. From the experimental results about discrimination and recognition, the following was obtained.
1) It was possible to discriminate fruit from its leaves and stems by use of both 550 and 850 nm wavelength interference filters.
2) It was effective for fruit discrimination to add pixels of high gray level in the image through 850 nm wavelength, when part of fruit had yellow green color or much bright reflection.
3) Fruit was almost recognized successfully by using features in the binary image and line image such as area, Feret's diameter ratio, continuity of horizontal intersectional line, centerline length and centerline direction.
4) Connection between centerlines considering the directions made fruit recognition easy and precise, even when fruit was divided by leaves or stems in the image.
