Intelligent Adaptive Algorithm for Boundary Detection of Dynamically Changing Regions in Video Sequences
Abstract
The problem of detecting the boundaries of dynamically changing regions in video sequences, such as flames and smoke, plays a crucial role in safety and monitoring systems. Existing edge detection methods (Sobel, Canny, Prewitt) are often insensitive to brightness variations in real-time video frames, leading to discontinuous or distorted contours. To overcome these limitations, this study proposes an intelligent auto-adaptive differential approach. The method is based on evaluating gradient variations, reducing noise using a Gaussian filter, and automatically adjusting threshold parameters through the Least Mean Square (LMS) algorithm. For each video frame, the algorithm adaptively updates the upper (TH) and lower (TL) thresholds to generate a continuous contour map. As a result, the proposed approach enables more accurate and smoother detection of dynamic object boundaries while significantly reducing background interference. Experimental results confirm the applicability of the method for real-time video analysis and its potential use in intelligent fire monitoring systems.