基于机器视觉和PNN神经网络的鸭蛋表面缺陷检测研究

Research on Surface Defect Detection of Duck Eggs based on PNN Neural Network

本章提出了一种基于概率神经网络(PNN)结合机器视觉的鸭蛋表面裂痕检测方法,配合背景光照法,结合灰度图像处理、图像高斯滤波处理、图像分割处理等算法去除图像杂质干扰,采用反锐化掩模局部对比度增强的分段增益改进算法来增强裂痕,收集裂痕、污点的相关信息作为数据集录入PNN神经网络,进行识别判断。本研究对1600张鸭蛋图片样本进行采样分析,将鸭蛋分为好蛋、脏污蛋、裂纹蛋3种。试验表明,该系统对干净无损蛋、脏污无损蛋、裂纹蛋的检测准确率分别达到了95.1%、77.9%、95.3%,具有较好的泛化性和鲁棒性,符合复杂鸭蛋生产加工环境的应用需求。

This chapter proposes a method for detecting cracks on the surface of duck eggs based on probabilistic neural network(PNN)combined with machine vision,combined with the background illumination method,combined with grayscale image processing,image Gaussian filter processing,image segmentation processing and other algorithms to remove image impurities.The segmental gain improvement algorithm for local contrast enhancement of sharpening mask is used to enhance cracks,and relevant information of cracks and stains is collected as a data set and entered into PNN neural network for identification and judgment.The study sampled and analyzed 324 image samples of duck eggs,and divided duck eggs into three types:good eggs,dirty eggs,and cracked eggs.Experiments show that the detection accuracy of the system for clean eggs,dirty eggs,and cracked eggs has reached 95.11%,77.9%,and 93.3%,respectively.It has good generalization and robustness,and is in line with complex duck egg production and processing.application requirements of the environment.