一种基于显著性检测的LED仪表字符自动识别方法

A Method of Automatically Recognizing Characters of LED Meters Based on Saliency Detection

针对LED数显仪表的字符显示值,提出了一种基于显著性检测的字符自动定位和识别方法。该方法利用LED图像的字符区域和背景区域在亮度和颜色上有明显区别的特点,通过提取图像中的显著性区域,并结合二值化和投影法,实现对字符区域的自动定位。再利用LED字符各段数码管之间的相对布局信息,可以准确识别每个字符。实验结果表明,本识别算法的识别率高达99.7%,同时,算法识别耗时很少,能够满足仪表字符识别的准确率和实时性要求。

Aiming at the display values of LED meters, we propose an automatic character localization and recognition method based on salient region detection. This method utilizes the distinctive differences in brightness and color between the salient character regions and background regions of the LED image. First, we evaluate the saliency map given the LED image. Then an automatic localization of the character regions is achieved via a binarization and projection method. Finally, the value corresponding to each character of the LED can be accurately identified by using the relative layout information between each segment of the LED characters. The comprehensive experiments show that the recognition precision is up to 99.7%, and the recognition takes very little time, which could satisfy the accuracy and real-time requirements for the recognition of LED meters.