Fruit harvesting robots have been emerging with the advancement of robotics and sensing systems. New methods of harvesting sweet cherries are being developed using mechanical branch shaking. Such techniques could be fully utilized with autonomous robotic harvesters. Autonomous harvesters should be able to locate tree branches and proper shaking locations for harvesting cherries. A computer vision based system is developed in this research to guide the robotic harvester to the proper shaking locations at the tree branches. Such computer vision-guided harvester will make the harvesters more efficient and will need minimal human interaction in making harvesting decisions.