Agriculture Reference
In-Depth Information
macropositioning platform, pneumatic power supply (end effector), and backup fuel
supply tank. The viewing window provides the operator/researcher a firsthand view
of the actual harvesting process.
7.4.3.1 Harvesting Manipulator
The manipulator used in this study is a Robotics Research Corp. model 1207 seven-
DOF manipulator. Figure 7.33 shows the manipulator and its kinematic model. It has
a maximum reach of 50 in, and its end-effector's velocity can reach up to 20 in/s.
The maximum payload capacity is approximately 20 lb. Similar to the human arm,
it has a shoulder, elbow, and wrist where the fruit gripping end effector is attached.
Although this manipulator was not specifically designed for fruit harvesting, it pro-
vided a high-precision manipulator that could be used for developing the other tech-
nologies. This 7-DOF manipulator is mounted on a 2-DOF platform that allows the
manipulator to move vertically and horizontally into the canopy. The added DOF
increase the working space of the manipulator, allowing the robot to harvest fruits
within the tree canopy.
7.4.3.2 Vision Sensory System
The vision system detects the fruit and localizes its position. It then guides the end
effector toward the fruit using visual servo control strategies. The vision system is
composed of a color CCD camera for image acquisition and a rack mount PC for
image processing. The CCD camera is connected to the PC using a frame grabber
that digitizes the image to a 640
480 24-bitmap image running at 30 frames/s. The
images are acquired under natural lighting condition.
There are numerous fruit recognition approaches that can be adopted based on vari-
ous features, which include (1) shape, (2) size, (3) spectral properties, or (4) texture.
In this study, the primary image processing steps are (1) segmentation by spectral
characteristics using an adaptive thresholding approach, (2) blob analysis, (3) circle
detection, and (4) centroid detection, which estimates fruit center position within the
image. Segmentation separated the fruit pixel from the background pixel by using the
color difference values with an adaptive thresholding approach. The color difference
×
Elbow
4
5
Elbow
Hand
3
7
Shoulder
6
2
1
Wrist
Wrist
Shoulder
Hand
Manipulator
Kinematics model of manipulator
FIGURE 7.33
Articulated manipulator for citrus harvesting (Robotics Research Corp.).
