Intelligent Perception Frameworks and Algorithms for Social Good: Food Security and Public Safety
As the global community faces increasingly complex challenges, the role of intelligent perception systems in addressing critical issues such as food insecurity and public danger has become more significant. By leveraging advancements in computer vision and machine learning/deep learning, these systems can process large volumes of sensory data and offer automated solutions that enhance decision-making and promote social good. This dissertation presents three innovative research works that design and apply intelligent perception frameworks and algorithms to tackle pressing global challenges in the fields of agronomy and public safety.
The first project introduces a probabilistic voxel carving (PVC) algorithm for reconstructing 3D models of maize plants, enabling the efficient extraction of key phenotypic traits such as leaf-stalk angles and leaf numbers. This research advances the field of plant phenotyping by providing an automated tool to support plant breeding and increase corn yield.
The second project focuses on soybean yield estimation, utilizing high-throughput seed counting in field environments. By applying a crowd-counting method and various strategies to improve model training, this work improves the accuracy and efficiency of seed detection and yield analysis, offering a scalable solution for crop breeding programs and advancing efforts to ensure food security.
The third project addresses public safety by developing a robust deep learning system for active shooter detection and tracking. Leveraging domain randomization and detecting both shooters and weapons, the system enhances the ability to identify and track active shooters in real time, contributing to more effective security infrastructure in public spaces.
Collectively, these research projects contribute novel methodologies in computer vision and machine learning, advancing the state-of-the-art in 3D plant reconstruction, agricultural yield estimation, and public threat detection. The research enhances the efficiency and accuracy of automated perception systems, demonstrating the potential of intelligent perception frameworks and algorithms to address real-world challenges. By improving agricultural productivity and advancing public safety measures, this work not only pushes forward the boundaries of technology but also has a meaningful impact on societal well-being.
Committee: Soumik Sarkar (major professor), Pavan Aduri, Baskar Ganapathysubramanian, Aditya Balu, and Adarsh Krishnamurthy
Join on Zoom: https://iastate.zoom.us/j/93715498193