Machine Vision System

Machine vision system was installed at a height of 0.6 m from the harvesting container (fig. 2). The focal length of the lens is 2.5 mm, to acquire the image of not only the fruits in the harvesting container, but also the fruits on the conveyer. If the machine vision system detects a fruit around the placement point on the conveyer, the unit stops until the fruit has  been conveyed by approximately 0.2 m from that point to maintain the separation between the supplied fruits. For the supply unit, the suction point should be the highest part of each fruit. Based on the recognition technique of tomatoes using specular reflection (Ota et al.,2010), we devised an algorithm for detecting the suction point of strawberry fruits by means of machine vision. If every suction point of all fruits has to be obtained from one shot image, local minima determination by Mathanker et al.(2010) could be a strong solution. The computation time for searching local minima points, however, would be relatively long as they indicated in the literature.Considering the supply unit handles only one fruit in a picking motion and grabs an image before each the motion,it should be enough for the supply unit to extract some candidates of the suction point from one shot image. Figure 4 shows a flow chart for extracting the suction point and figure 5 shows an example of detected candidates of the reflecting part.

An image with VGA resolution is acquired by a color CMOS camera. The region of the harvesting container is cropped. The reduced image is then broken down into RGB gray images and transformed into HSV coordinates. The V coordinate is binarized using the discrimination method(Otsu, 1979) to obtain the fruit regions. Although the double Otsu method of Mathanker et al. (2010) can determine two thresholds: one for the fruit regions and another for the bright regions, it is estimated that the bright regions in relatively dark fruit regions will be lost using the common threshold for all the bright regions, and the number of candidates for the suction points will be reduced.Thus, in each of the extracted fruit regions, 10% of the bright area is segmented using the percentile method. We assumed that the bright regions were the reflecting parts,

namely, the highest part of the fruits. After a labeling and morphology process to eliminate noise, the points of gravity of the reflecting parts are calculated. The nearest fruit to the left inner side wall of the harvesting container in figure 5a has priority for being picked up. The supply unit repeats the task until the harvesting container is empty. However, to prevent multiple attempts on the same failed fruit that might damage it, the unit records the suction point and tries to pick up the fruit only twice if the fruit is in a circle within a 10-mm radius of the logged points. The computation time required less than 0.3 s for the image processing.

Figure 3. Schematic diagram of the suction hand of the supply unit.

Figure 4. Flowchart of extracting suction point.

Figure 5. Detection of suction points on fruits with machine vision system

PACKING UNIT

Hayashi et al. (2011) have developed strawberry pickup equipment that detects the position and orientation of the fruit using machine vision system, and picks up the fruit from the calyx side with the suction device. We have added new functions to the equipment such as the packing unit

(fig. 6) of the automatic packing system.

Placing Motion

The packing unit’s suction device is equipped with a step motor which tilts the suction tube between -90°and+90°. When the suction device picks up the fruit, it tilts the suction tube at 10°from the horizontal. The tilt angle of the suction tube changes to vertical when the  packing

unit transports a fruit, to optimize stability of transportation, as reported by Hayashi et al. (2011). The fruit is packed straight into the returnable tray vertically as shown in figure 7a. Meanwhile, the tilt angle of the suction tube is set at 10° again when the fruit is packed in the single-layer tray (fig. 7b).