Research On Flexible Object Grasping Method Based on Visual Enhancement and Multi-Stage Collaboration

To address the challenge of grasping reflective, low-texture flexible objects, this paper proposes a framework combining visual enhancement and multi-stage collaboration. It improves image quality, enables accurate “unfolding–grasping” transitions, and detects grasp points in real time via instance segmentation and coordinate mapping. Experiments show 35% and 26% improvements in transition and grasp success rates over traditional methods, providing an effective solution for service robots.