MetadataShow full item record
PublisherThe University of Arizona.
RightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
AbstractA GPU-accelerated real-time CGH computation and display framework for beam steering is reported in this dissertation. This approach utilizes CUDA for CGH computation and OpenGL API for 2D CGH rendering. The deterministic method for CGH beam steering is suitable for GPU computation because such method possesses low complexity, massive calculation, and independency of each pixel, which enables pixel-wise parallel processing with GPU. The performance of the presented approach for GPU-based CGH calculation and display on GTX 1650-Ti, results in ~10 times faster than CPU-based processing on i7-10750H@2.6GHz, which also achieves the maximum refresh rate (180 FPS) of the current generation of Texas Instrument Phase Light Modulator (PLM). In addition, we combine deep learning model (YOLOv4-tiny) for object detection and recognition, enable the system doing AI-based dynamic beam steering and trace the object of interest. Instead of single beam steering, multi-beam with variable beam ratio beam steering is also conducted on the same GPU framework, resulting in the beam steering speed more than 1k beams/second. Furthermore, the beam ratio simulation is introduced, which matches the measured beam ratio results. This simulation model allows us to predict and control the energy distribution. Thus, the AI-based dynamic multi-beam tracking with varying beam ratio is demonstrated in this dissertation
Degree ProgramGraduate College