Robotic System to Identify Finite Number of Significant Image Frames on Large Video Data using Unsupervised Machine Learning Technique

In this research work, a video-related automated system, namely Robotic Key Image Frame Identification System (RKIFIS), is proposed, and it aims to instinctively identify the finite number of representative image frames over the video through the process of splitting the video contents into the optimal number of distinct clusters with different sizes using the Optimal-N-Means ONM clustering technique. The proposed RKIFI system contains five stages; in the beginning stage, the RKIFI converts the input video into a sequence of image frames using the standard open CV tool. Subsequently, the proposed system improves the image frame quality through pre-processing every individual image frame from the result of the previous stage. Afterward, the RKIFI system extracts highly relevant features from each image frame in the image frame set of the input video using standard arithmetic operations. Consecutively, the proposed system is iteratively split into the image frame vector set into a finite number of clusters through the process of iteratively identifying the optimal number of representative image frames over the input image frame set of the input video using the Optimal-N-Means clustering technique, where N denotes the optimal number of representative image frames in the image feature vector set of the input video. In the final stage, the RKIFI system validates the dissimilarity level among the key image frames which are identified in the clustering stage. The experimental result shows how the RKIFI system is well suited to automatically identifying the essential key image frames in the video data.