Recent Inspection Techniques on Imaging and Data Science for Complex Data Analysis

 
Recently, there has been an increasing demand to develop inspection techniques on imaging and data science for complex data analysis. Inspection is a way to detect, identify and locate any possible anomalies in any process of manufacturing, imaging, data transmission and intelligent systems. In this talk, I will present some recent developed techniques. Patterned texture inspection is one popular topic in imaging science, especially for manufacturing and printing industries. The recent methods like Bollinger bands and regularity analysis and the latest methods like Elo rating method, image decomposition model, total variational and sparsity with low rank model will be discussed. Next, examples of utilizing computer vision techniques to discover social groups in a protest event in visual surveillance and to register the physical and computerized domains of a cone-beam computed tomogram for a dental implant surgery in medical imaging will be given. In data science, modern cities experience heavy traffic flows and congestions regularly across space and time that is a typical problem in big data analytics. Monitoring traffic situations becomes an important challenge for the Traffic Control and Surveillance Systems (TCSS). In advanced TCSS, it is helpful to automatically detect and classify different outliers such as data errors and traffic incidents such as severity of congestion, abnormal driving pattern, abrupt or illegal stop on road, etc. In literature, many outlier detection (OD) methods are available but there is a lack of research for Automated Incident Classification (AIC). I will introduce the recent OD methods such as k-nearest neighbor, local outlier factor, quaternion-valued model, Dirichlet process mixture model, etc. and the recent AIC methods such as adaptive boosting support vector machine, modulo-k clustering tree methods based on a large-scale traffic database.