LiDAR technology is pivotal in various applications, such as autonomous vehicles, remote sensing, and 3D mapping.
This study aims to explore the computational capabilities of Central Processing Units (CPUs) and Graphics Processing Units (GPUs) for LiDAR point cloud data processing, considering factors like data size, algorithms, and hardware architectures. The research involves developing CPU and GPU-based implementations for common LiDAR data processing tasks, including point cloud filtering, segmentation, and feature extraction, using real LiDAR datasets of varying sizes for benchmarking.
By comparing execution times, throughput, and resource utilization, the study demonstrates the significant performance advantages of GPUs over CPUs for LiDAR data processing, consistently showcasing advanced processing speeds and throughput. Furthermore, the findings highlight trade-offs, challenges, and considerations when choosing between CPU and GPU platforms for LiDAR data processing, ultimately pointing towards the potential of GPU-accelerated solutions for real-time and efficient LiDAR data analysis across diverse industries.
This project aims to gain a comprehensive understanding of multi-core microcontroller architecture and programming by developing an application to measure distance with an ultrasonic sensor and indicate the proximity to an object using LEDs. The hardware platform chosen for this project is the Aurix TC375 board, which features multiple cores that can independently handle different peripherals. GPIOs (General Purpose Input/Output) are used to interface with the ultrasonic sensor and control the LEDs.
The project focuses on the following key aspects:
- Multi-Core Processor Architecture
- Ultrasonic Distance Measurement
- Proximity Indication
- Independent Core Utilization
- Intelligent Controls --> Transfer Learning --> Image classification using Alexnet
This challenge is to build your own Neural Network in MATLAB using transfer learning technique to modify AlexNet ---> KiranNet. The task is to train a CNN to recognize 5 different types of objects.
- Introduction to MATLAB & Model Based Design - Onboarding training for new joinees