Page Banner

United States Department of Agriculture

Agricultural Research Service

Related Topics


Location: Sugarbeet and Bean Research

2012 Annual Report

1a. Objectives (from AD-416):
The proposed research is to develop technology for presorting apples in the orchard. Specific objectives are to: 1) develop a cost-effective machine vision-based grading unit that can sort apples into two or three grades (cull, processing, and fresh market); 2) develop a fruit handling unit for loading, transporting, unloading the graded fruit into individual fruit bins; and 3) assemble and evaluate a presorting prototype that is mobile, rugged, and easy to operate in the orchard.

1b. Approach (from AD-416):
Color vision technology will be applied to develop an in-orchard apple grading station which will sort apples into two or three grades (cull, processing, and fresh market). A laboratory machine vision unit will be built and tested. This vision unit will be designed to accommodate a harvest crew of 8 to 10 people. Algorithms will be developed to acquire and process apple images and segregate them into different grades. Our initial goal is to sort and grade fruit based on their size. Additional capabilities such as sorting for color and blemish will be considered at a later time. In the mean time, we will compare different fruit handling designs and select and build a new fruit handling unit that is suitable for use in the orchard. The handling unit will allow fruit pickers to unload the fruit in the picking bags, transport to the grading station for imaging and grading, and then place the graded fruit into appropriate fruit bins. Once the imaging unit and fruit handling unit are built and tested, we will then assemble an in-orchard presorting prototype. This prototype will be tested in a commercial orchard. Finally, a field demonstration to growers will be arranged.

3. Progress Report:
Research for the project in the past year was primarily focused on designing and assembling a new version mobile system for infield sorting of apples, with the additional effort being made to improve and expand the fruit sorting algorithm. Based on the test results for the earlier version presorting system, special considerations were given in the design of several key components for the system, including fruit orientation and rotation for imaging, fruit conveying and bin filling. With these considerations, a new version presorting system was designed and assembled. The system uses a unique fruit singulation and conveying design that is easy to assemble and low in cost. The new bin filler design also is simpler and more compact compared to other existing designs. The new system can sort apples into two or three quality grades to meet the varying needs from different growers. By using an existing harvest trailer or a similar harvest machine, it significantly reduces the unit machinery cost. This presorting system can accommodate a harvest crew of six to eight people; it also incorporates some harvest aid functions, so that fruit pickers can pick fruit on the trees from the ground or the platform without using ladders, which improves the working condition and safety for workers. In parallel with the sorting system development, two separate studies were carried out to evaluate the effect of chemical residue and dust at the surface of apples on the color grading performance, and to optimize the image processing algorithm to improve color grading accuracy. The residue/dust study for ‘Delicious’ apples showed that the presence of chemical residue or dust on the surface of apples did not have a significant effect on fruit color classification. The color imaging study suggested that the color grading results could be improved when using a fewer select images for each fruit instead of all images (~16 images for each fruit). This finding is expected to significantly improve the image processing speed for color and defect sorting and grading. In addition, color images were collected from defective apples (i.e., scab, cuts, bird damage, disease or insect damage, misshape, and other surface blemishes) using a laboratory machine vision system to build a comprehensive database for different types of defect. Initial results showed that the algorithm is effective in automatic detection of more visible defects. Current effort is being focused on detecting those defects that are less distinct visibly. Once completed, the defects detection algorithm will be incorporated into the new presorting system.

4. Accomplishments

Last Modified: 2/23/2016
Footer Content Back to Top of Page