图像识别技术研究有色金属大气腐蚀早期行为

ATMOSPHERIC FOREPART CORROSION BEHAVIORS OF NONFERROUS METAL BASED ON IMAGE RECOGNITION

提出了基于图像识别技术研究有色金属大气腐蚀早期行为的方法,用扫描仪获取试样的大气腐蚀形貌图像,采用连续小波变换对图像进行分解并提取能量值作为特征信息,研究了图像特征值和试样腐蚀失重数据之间的相关性.运用该方法分析了北京郊区纯锌的大气腐蚀试样.结果表明,该方法可以对纯锌的大气腐蚀早期行为进行判断和预测.

An analyzing method based on image recognition is set up to study the atmospheric forepart corrosion behaviors of nonferrous metal: first, scanner is used to acquire corrosion images of nonferrous metal in atmosphere environment; second, multi-resolution continuous wavelet transformation is applied to decompose the images and energies of sub-images are extracted as character information; finally, the relationship between the characters and corrosion data of samples is studied by peculiar way, such as artificial neutral network, regression model, etc. Using the established method, the pure zinc samples in Beijing atmosphere are examined and a model, which has a high precision, 0.98 and can predict the corrosion weight lose data of samples is erected. Furthermore, another method that can be used to forecast the forepart corrosion degree of pure zinc in atmosphere was primary studied, too.