Abstract: In granule processing industries，acquisition of particle size and shape parameters is a common procedure，and volumetric measurement is of great importance in dealing with particle sizing.To eradicate the major drawbacks with manual gauge，this paper proposes an optical approach using BP neural network to estimate the particle volume based on the 2-D image information.To achieve the better network efficiency and structure simplicity，Principal Component Analysis（PCA） is adopted to reduce the dimensions of network inputs.To improve the network convergence，momentum is incorporated in gradient descent algorithm.The real particle data are applied in training and testing the presented network.The experiment results suggest that the proposed neural network is capable of estimating particle volume with satisfactory precision and superior to the conventional BP network in terms of performance capacity.

Key words: particle image, particle parameters, Principal Component Analysis（PCA）, neural network, volume estimation

摘要： 在颗粒加工工业中，获取颗粒尺寸和形状参数是一道常见的工序。体积是一个重要的颗粒三维参数，采用传统的手工测量方法获取体积耗时长，人工投入较多，很难实现过程控制中的实时反馈。应用计算机视觉技术，提出了一种基于颗粒单视二维图像信息（周长、投影面积、长宽比等）的BP神经网络体积估算方法。为了避免传统BP神经网络收敛速度慢，容易陷入局部极小值的缺陷，采用BP神经网络的改进算法-有动量的梯度下降算法。同时应用主成分分析法来进行体积影响参数的降维处理，减小了网络结构复杂度并提高了网络的整体性能。使用真实颗粒图像及实测数据对神经网络进行训练和精度测试，结果表明，将主元分析法与BP神经网络相结合来进行体积估算无论在预测精度还是在网络运算速度上比全要素传统BP神经网络模型具有更大的优越性。

关键词: 单视二维图像, 特征参数, 主元分析, 神经网络, 体积估算