中文摘要：

金属断口图像中标定符号信息是进一步计算图像对应实际物理空间距离的依据。标定符号通常为印刷体,所以准确定位是正确识别的前提和关键。对强噪声复杂背景下的金属断口图像标定符号的定位算法的研究,先对直线特征明显的标尺符号定位,其中对Radon变换进行分块改进,使快速性和准确性有了明显改善。字符定位利用符号的纹理特征进行数学形态学粗定位和图像边缘模板匹配精定位结合的方法,并根据标尺位置和长度等信息缩小搜索区域。实验结果表明,该算法的定位准确率达到94%。

英文摘要：

The information of calibrated symbols in metal fracture images is the foundation of further obtaining physical spatial distance the image corresponded. Calibrated symbols are usually printed fonts, therefore the accurate location is the precondition and key of the recognition. The algorithm of calibrated symbol location of metal fracture image in high noisy background is studied in this article. The straight line feature of ruler symbols is obvious,thus its processing is prior to character location. Traditional Radon transform is improved based on block to realise faster and more accurate performance. The character location based on texture feature is carried in two steps, rough location by mathematical morphology and accurate location by image edge template matching. The searching region is reduced by combining the position and length of ruler symbols together. The accuracy rate reaches 94% as shown by experiment results.