Computer Vision for Detecting Invisible and Transparent Objects Using Motion Distortion

Invisible and transparent objects such as glass doors, plastic sheets, water spills, and transparent containers pose a significant challenge to computer vision systems. Traditional object detection models fail due to minimal texture, low color variance, and absence of edges. This paper proposes a novel motion-distortion-based detection framework that analyzes microvariations in optical flow, refractive distortions, pixel jitter, and background displacement to identify transparent or invisible objects in video sequences. The model integrates a dual-path pipeline: (1) Refractive Flow Estimation (RFE) and (2) Motion-Distortion Residual Mapping (MDRM). Experimental evaluation on custom video datasets demonstrates successful detection of glass panes, plastic sheets, and water surfaces with 87.4% accuracy, outperforming baseline edge-based and YOLO-based models. This research highlights a promising direction for safer robotics, autonomous navigation, forensic analysis, and assistive systems, especially in indoor environments.