Scholarlink

This work presents a novel Adaptive backpropagation neural network (ABPNN) algorithm using Neuron dependent Sigmoid function (NDS) and Neuron dependent weight updation rule (NDW) for three dimensional Light detection and ranging (3D LiDAR) point cloud (PCD) compression. This compression algorithm consists of two processes. The raw PCD data is downsampled and normalized by applying the Nyquist-Shannon sampling and Min-Max normalization methods to enhance the PCD data. Then, the enhanced data quantized by learning ABPNN algorithm using the proposed NDS and NDW. The proposed ABPNN is outperforming well the two different combinations of the backpropagation and polynomial algorithm. The experimental result shows that this proposed algorithm averagely increases the prediction accuracy by 33.4% more than the existing Sigmoid function and it produces minimum constant compressed bitstream with better quality PCD within less time. Hence, this ABPNN algorithm using NDS and NDW is the best-suited method for 3D LiDAR PCD spatial information compression.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive Backpropagation Algorithm Using N-Sigmoid and N-Weight Updation Rule for 3D LiDAR Point Cloud Compression

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • A. Christoper Tamilmathi,
  • G. Dheepa

摘要

This work presents a novel Adaptive backpropagation neural network (ABPNN) algorithm using Neuron dependent Sigmoid function (NDS) and Neuron dependent weight updation rule (NDW) for three dimensional Light detection and ranging (3D LiDAR) point cloud (PCD) compression. This compression algorithm consists of two processes. The raw PCD data is downsampled and normalized by applying the Nyquist-Shannon sampling and Min-Max normalization methods to enhance the PCD data. Then, the enhanced data quantized by learning ABPNN algorithm using the proposed NDS and NDW. The proposed ABPNN is outperforming well the two different combinations of the backpropagation and polynomial algorithm. The experimental result shows that this proposed algorithm averagely increases the prediction accuracy by 33.4% more than the existing Sigmoid function and it produces minimum constant compressed bitstream with better quality PCD within less time. Hence, this ABPNN algorithm using NDS and NDW is the best-suited method for 3D LiDAR PCD spatial information compression.