复杂光照下变电站指针式仪表图像自动识别研究

Automatic Recognition of Pointer Instrument Image in Substation Under Complex Illumination

在无人巡检变电站中光照环境复杂,巡检机器人拍摄的照片总会出现光照不均匀现象,为消除复杂光照对图像识别的影响,首先使用基于Retinex理论的自适应Gamma增强算法去掉仪表图像不均匀光照的影响,使图像亮度均衡,然后使用双阈值算法对图像二值化,再进行骨架化处理,用Hough变换对指针提取,最后用查表法读取仪表示数。试验表明,使用该算法能有效克服不均匀光照的影响,二值化效果良好,指针仪表示数读取准确率提高。

The illumination environment is complex in unattended substation inspection. In the process of inspection, the photos taken by the robot always appear uneven illumination. A solution was proposed to eliminate the complex illumination for image recognition. Firstly, an adaptive enhancement algorithm based on Gamma Retinex theory was used to remove the nonuniform illumination of meter image and make the image brightness equalization. Secondly, the double threshold algorithm was used to image binarization and then processed to extract the skeleton. The pointer was extracted with Hough transform and finally the look-up table method was used to show the number of reads. Experiments showed that the algorithm could effectively overcome the influence of uneven illumination, and has good binarization value. The accuracy of pointer meter reading accuracy was improved.