采用图像识别技术对动车组零部件用S355J2W钢板的带状组织进行评定。通过摄像头从显微镜中获取试样显微组织的图像，采用数字图像处理方法对图像进行处理，对图像进行特征物提取和特征物计数，最后计算出评定结果。结果表明：S355J2W钢板带状组织评定结果相对精度小于30%，符合带状组织评定标准要求；采用图像识别技术减少了主观因素的影响，提高了工作效率。

Banded structure of S355J2W steel plate for EMU parts was evaluated by image recognition technology. The microstructure image of the sample was obtained from the microscope by the camera. The image was processed by digital image processing method, and then the feature objects were extracted and counted. Finally, the evaluation result was calculated. The results show that the relative accuracy of banded structure evaluation of S355J2W steel plate was less than 30%, which met the requirements of banded structure evaluation standard. The influence of subjective factors were reduced and the work efficiency was improved by the image recognition technology.