The utilization of real time image analysis within the IoT networks has received increased interest because of the high demand of visual information that will support applications like smart cities, industries, and autonomous structure. In this paper, presents a reliable a statistical-AI combined model for real-time image processing in an IoT framework. The proposed system is based on edge computing in order to minimize the amount of time needed to process and enhance the alterations, as well as to reduce the burden on cloud servers. CNN is used for Image classification and Sobel operator is used for extracting the edges of the images to augment feature extraction. Experimental outcomes show 50% less time for processing when using the concepts of edge-based systems although the classification accuracy increases from 85% to 92% as the epoch of training is enhanced. The results prove that the proposed system is scalable for that the performance is good even under different IoT network topologies, hence suitable for different real time applications.