中文摘要：

采用基于机器视觉的检测系统替代人工检测方式，对发动机气门电镦成型加工过程中气门棒料人模状态进行检测。通过对气门棒料入模状态的图像处理分析，检测气门棒料是否正常人模，以提高发动机气门生产的合格率。对比分析了灰度平均值法、灰度腐蚀以及空域和频域下模板匹配这四种可行视觉检测方案的实验结果，证明了灰度平均值视觉检测方案具有良好的时效性与准确性，能够满足发动机气门电镦成型对气门入模状态检测的性能要求。

英文摘要：

The system of detecting the valve status into die is based on machine vision instead of manual test mode. The principle is that detect whether the valve into die is correct through the result of processing and analyzing the gotten image of valve status. The method can improve the production yield. There are four viable visual detection methods which are average gray-value, gray scale corrosion, template match in area space and template match in frequency space. By comparing the four methods find that the method based on the average gray-value has good timeliness and accuracy which meet performance requirements of the detection.