Machine Vision System for Urban Scenic Detection

Optical flow, such as the Gunnar-Farneback algorithm, is used in traffic to find the displacement between two image frames. To find an anomaly in videos, many studies have used optical flow as a basis for surveillance analysis together with other technologies such as machine learning and deep learning. However, newer technologies require more data science expertise and more resources such as a labeled dataset. The proposed system utilizes the Gunnar-Farneback optical flow algorithm to detect displacement and explores the data analysis techniques by collecting the overall normalization values to flag an anomaly. The system is designed using personal computers, Python, and MATLAB, while other tools such as FFMPEG and Microsoft Excel are used for video segmentation and data representation. However, testing shows that there is more work needed to optimize the analysis technique for defining an anomaly in videos.