Abstract

In this article, we suggest a new approach for classification and Recognition of 3D image Gaussian–Hermite moments using a Multilayer Perceptron architecture. The Multilayer Perceptron is an artificial neural network to evaluate the efficient structure in the non-linear systems. However, the determination of its architecture and weights is a fundamental issue due to their direct impact on the network convergence and performance. The robustness of the proposed approach have provided under many transforms. The experimental results show that our approaches are more robust than 3D Geometric moments.

中文翻译：

---

基于不变高斯-赫姆特矩的神​​经网络用于3D对象分类

摘要

在本文中，我们提出了一种使用多层感知器体系结构对3D图像高斯-赫姆特矩进行分类和识别的新方法。多层感知器是一种人工神经网络，用于评估非线性系统中的有效结构。但是，确定其架构和权重是一个基本问题，因为它们直接影响网络的融合和性能。所提出的方法的鲁棒性已经在许多变换下提供了。实验结果表明，我们的方法比3D几何矩更可靠。