基于光度立体法的汽车零件字符识别预处理研究

Research on Character Recognition Preprocessing of Automobile Parts Based on Photometric Stereo Method

为了提高工业自动化对汽车零件上字符的自动识别效率,论文针对汽车零件上字符颜色一般与背景色相近,两者对比度非常低的问题,提出了基于光度立体法的汽车零件字符预处理方法。首先从四个不同方向对物体进行照明收集图片,利用四张采集的图片计算物体三维表面的局部梯度,然后继续计算物体表面的旋度、散度、平均曲率和高斯曲率,最终得到对比度高的灰度图片,提高后面字符识别过程的效率。

In order to improve the automatic recognition efficiency of characters on auto parts in industrial automation,this pa⁃per proposes a method for character processing of auto parts based on photometric stereo method for the problem that the color of characters on auto parts is generally close to the background color and the contrast between them is very low.Firstly,the objects are illuminated from four different directions to collect images,and the four collected images are used to calculate the local gradient of the three-dimensional surface of the object,and then the curvature,divergence,mean curvature and Gaussian curvature of the sur⁃face of the object are continuously calculated,and the contrast is high finally.Grayscale images improve the efficiency of the charac⁃ter recognition process.