National Institute of Technology, Karnataka

About: National Institute of Technology, Karnataka is a education organization based out in Mangalore, Karnataka, India. It is known for research contribution in the topics: Computer science & Corrosion. The organization has 5017 authors who have published 7057 publications receiving 70367 citations.

Determination of Wetting Behavior, Spread Activation Energy, and Quench Severity of Bioquenchants

Abstract: An investigation was conducted to study the suitability of vegetable oils such as sunflower, coconut, groundnut, castor, cashewnut shell (CNS), and palm oils as quench media (bioquenchants) for industrial heat treatment by assessing their wetting behavior and severity of quenching. The relaxation of contact angle was sharp during the initial stages, and it became gradual as the system approached equilibrium. The equilibrium contact angle decreased with increase in the temperature of the substrate and decrease in the viscosity of the quench medium. A comparison of the relaxation of the contact angle at various temperatures indicated the significant difference in spreading of oils having varying viscosity. The spread activation energy was determined using the Arrhenius type of equation. Oils with higher viscosity resulted in lower cooling rates. The quench severity of various oil media was determined by estimating heat-transfer coefficients using the lumped capacitance method. Activation energy for spreading determined using the wetting behavior of oils at various temperatures was in good agreement with the severity of quenching assessed by cooling curve analysis. A high quench severity is associated with oils having low spread activation energy.

Dynamic video anomaly detection and localization using sparse denoising autoencoders

Abstract: The emergence of novel techniques for automatic anomaly detection in surveillance videos has significantly reduced the burden of manual processing of large, continuous video streams. However, existing anomaly detection systems suffer from a high false-positive rate and also, are not real-time, which makes them practically redundant. Furthermore, their predefined feature selection techniques limit their application to specific cases. To overcome these shortcomings, a dynamic anomaly detection and localization system is proposed, which uses deep learning to automatically learn relevant features. In this technique, each video is represented as a group of cubic patches for identifying local and global anomalies. A unique sparse denoising autoencoder architecture is used, that significantly reduced the computation time and the number of false positives in frame-level anomaly detection by more than 2.5%. Experimental analysis on two benchmark data sets - UMN dataset and UCSD Pedestrian dataset, show that our algorithm outperforms the state-of-the-art models in terms of false positive rate, while also showing a significant reduction in computation time.