ABSTRACT
Precision and flexibility are required from robotic pick-and-place actions in a variety of settings; especially when devices take photographs with different characteristics. This represents a challenge for conventional deep models, reducing their utility in more practical situations. This paper introduces the integration of HS-CLAHE with NFPFLU-ResNet50 with a purpose to develop an adaptive motion planning framework that would further facilitate object recognition, segmentation, and motion planning in achieving effective and resilient robotic automation in dynamic scenarios.It employs NFPFLU-ResNet50 for precise feature extraction and motion planning, ensuring excellent performance in noisy environments and HS-CLAHE for image preprocessing with better contrast across devices. This framework finds amazing improvement in terms of the accuracy up to 94% and precision up to 93% and recall of up to 92%. In addition, it manages to reduce the Root Mean Error up to 3.8%. In reality, this algorithm proves as efficient as traditional approaches regarding real-time pick-and-place task.In that regard, the adaptive framework, using up-to-date cutting edge deep learning techniques, along with enhancing quality pictures, managed to considerably improve motion planning in robots. It thus performed pretty well under an array of different operational settings, giving it a fairly relevant application in industrial automation applications.
