Multicore Real Time Feature Detection System Using Thermal Video for Nondestructive Testing

Multicore systems are widely used to improve performance and satisfy time and power demands. Indeed, multicore systems increase the speed of the execution by enabling an algorithm to run multiple tasks in parallel. In this work, we use thermal video captured for the purpose of nondestructive testing while aiming for automated real time analysis. Thermal videos for inspection purposes have been in the forefront of many applications such in quality control in manufacturing, in searching for living bodies in large fires, and in finding lost people in the wilderness. We aim with this work to develop a multicore system to speed up processing and analysis operations of thermal video for the purpose of feature detection and classification. Experimental results show that the processing time for Canny edge detection algorithm is improved 1.93 times using 2 processes, and 2.6 times using 4 processes while histogram equalization of the thermal video frames was improved 2.23 times using 2 processes, and 2.73 times using 4 processes.