MACHINE VISION SYSTEM FOR ROBOTIC APPLE HARVESTING IN FRUITING WALL ORCHARDS
MetadataShow full item record
Mechanization in agriculture is often regarded as one of the greatest human achievements in the twentieth century. These technological advancements have led to significant reduction in human effort in the production of bulk agricultural commodities such as corn and wheat. Specialty crop industry such as fresh fruit market, on the other hand, are still dependent upon manual labor for various production operations such as training, pruning, and harvesting. Among these, tree fruit harvesting of high value crops like apples is the most labor intensive and time sensitive task that requires the right number of farm workers at right time. To increase productivity and reduce dependency on seasonal labor, researchers have proposed automated harvesting systems. Because of highly unstructured orchard environment and variable outdoor conditions, these technologies have achieved only limited successes in the past. No commercial viability has been achieved and every apple destined for fresh market is still handpicked. The lack of mechanized harvesting system has the potential to threaten the long-term sustainability of fresh fruit industry in the United States and around the world. This dissertation focuses on the study and evaluation of a machine vision and an integrated robotic system for automated harvesting of apples grown in modern fruiting wall orchards. The machine vision algorithm designed to work in orchard environment accurately detected apples growing individually as well as in heavy clusters under variable natural lighting conditions. A pragmatic approach to harvesting (also called hierarchical approach) showed that 98% of the fruit could be detected with iterative imaging and harvesting of most visible fruits. The integrated robotic system with global camera and custom built seven degrees of freedom manipulator successfully picked 84% of attempted fruit with 6 seconds of average harvest time per fruit. This approach of selective apple harvesting with a global camera system and low-cost manipulator show a huge potential for cost-effective robotic solutions for harvesting fresh market apples. However, several limitations still remain to be addressed for commercialization.