论文部分内容阅读
研究了采用高斯滤波、模糊增强、阈值分割和特征提取等数字图像处理技术对采集的铸造缺陷图像进行处理,获得带有各种铸造缺陷特征的图像样本,并用BP神经网络对样本进行训练,最终获得具有较高识别精度的BP网络。实验结果表明,利用该BP网络可以实现对铸造缺陷图像类型的快速、准确判断。
The digital image processing techniques such as Gaussian filter, fuzzy enhancement, threshold segmentation and feature extraction were used to process the images of foundry defects, and the image samples with various casting defects were obtained. Finally, the BP neural network was used to train the samples. Finally, Obtain BP network with higher recognition accuracy. The experimental results show that the BP neural network can be used to quickly and accurately judge the image type of casting defects.