The first paper focuses on YOLOv5 model optimization for advanced image processing, cu. The second paper explores low-cost machine vision cameras, achieving 97% accuracy in grasp verification. The third paper implements SVM algorithm with ArUco markers for hand-eye calibration. The adjustments made to the connection of sensing and motion system significantly expanded the scope of the detection while maintain the accuracy of grasping.. The fourth and fifth experiments improved the existing models by adding coordinate systems and wrist-mounted cameras separately. By doing so, it is possible to catch clearer and more precise pictures of the target objects. The final two research focus on promoting the automatic aspect of the system. Utilizing two networks and corresponding datasets, for one; employing multiple innovative vision-based techniques, for another. These studies collectively demonstrate the advancement in machine vision-based robotic grasping through various technical approaches, including model optimization, hardware evaluation, and algorithm implementation for improved accuracy and efficiency.
Read full abstract