Chandra Kambhamettu, a Professor in the Department of Computer & Information Sciences, has been awarded a nearly $4 million four-year research grant from the Army Research Lab

The project, “A Comprehensive Multi-vehicle based Deep Learning System for Diurnal Multimodal Hidden Targets Detection,” will develop a robust, deep learning-based detection system for spotting hidden targets. Using multi-vehicular platforms to collect multi-spectral data, these algorithms will be tested and refined through field work at U.S. military bases

Kambhamettu’s research group, the Video/Image Modeling and Synthesis (VIMS) Lab, has been collaborating with the ARL on automated landmine detection since 2006. Previously, the researchers used data collected by ground vehicles and drones from color, thermal and infrared cameras along with GPS, Lidar and radar to spot above ground landmines at various test sites, including a recent trip to the Marine Corps Base Camp Pendleton in Oceanside, CA.

Now, in their latest project, which is one of the largest single investigator awards received by a UD computer science faculty member, the researchers will enhance their existing system by incorporating data from additional channels over a 24-hour time period. The long-term goal is to create a device that provides real-time detection results and visualization for decision-making in the field.

“Our role is very critical in that we receive all the volume of data and are the ones who actually make the decisions based on complex machine learning models,” Kambhamettu explained. “This approach in the field of computer vision, where we develop novel AI algorithms that can perform hidden target analysis, is very unique.”

Along with their unique approach to addressing the challenging problem of hidden target detection within the field of computer vision, this work could also have broad societal impact by providing real-time information on landmine locations to soldiers and ground troops in the future.

“Leveraging the data collected from multiple sensors on multiple vehicles, including drones, in an integrated way to make real-time decision with high confidence, is a key component to realize the vision of ‘Vehicle Computing’, and here at the University of Delaware we have a critical mass of faculty and students working on connected and sustainable mobility,” said CIS Department Chair Weisong Shi. “Professor Kambhamettu’s multimillion project on hidden target detection is the latest example demonstrating our excellence of research in computer vision and mobility.”