Cognitive Processing and Exploitation of 3D Laser Imaging Detection and Ranging (LIDAR) Imagery Data

This STTR Phase II project will develop a hierarchical visual processing system and attention-based mechanisms based on novel 3D convolutional neural networks to enable robust and efficient detection and classification of objects in 3D LIDAR imagery. The Phase II will build upon the enabling technologies developed in Phase I which, using deep learning methods, model the ventral pathway of the visual cortex and address many limitations of the state-of-the-art solutions that perform exhaustive search for the targets of interest and rely on pre-designed features. Our system uses bottom-up and top-down attentional mechanisms to improve the efficiency of the framework and enable incorporation of prior models and available context for efficient performance. Moreover, instead of using hard-coded features, the system is capable of learning hierarchical invariant 3D feature representations that are object class independent from training datasets. The Phase II effort will include improving the core technologies, optimizing and porting the technologies for efficient processing to handle large 3D LIDAR datasets, and demonstration of the technologies. The project will benefit from Novateur Teams expertise in the areas of biologically inspired cognitive systems, attention-based vision, 3D deep learning and convolutional neural networks as well as sensor exploitation.