Adaptive uneven illumination correction method for autonomous live-line maintenance robot

With the development of the robot in electricity, more and more autonomous live-line maintenance robots (ALMRs) have been developed and put into use. However, in outdoor environment, complicated and uneven illumination lead to huge challenges for visual feedback and target recognition of ALMRs. Aiming at easing the disturbance brought by the strong uneven illumination, we collect a Hot-Line dataset containing fieldwork photos of the ALMR and propose an image enhancement method for uneven illumination images based on image brightness segmentation and multi-methods fusion. Through image segmentation based on illuminance, the proposed algorithm enhances the over- and under-illuminated parts of the image differently while taking the approximate illumination component as a reference. We introduce an adaptive weighted summation strategy to ease the problem of edge transition in the output. The proposed algorithm improves the overall performance of a fieldwork image of ALMR properly, making the image clearer and better. For six indexes (Laplacian, SMD2, Energy of Gradient (EOG), and Entropy for image clarity; Structural similarity index measure (SSIM) and Peak signal-to-noise ratio (PSNR) for the degree of image information retention), the proposed method provided good results on both our Hot-Line dataset (for example, on EOG, the proposed method achieves nearly double the performance index value than CLAHE) and other image datasets, and finished the enhancement within a relatively short time (within 0.02s with image size 275 x 275). The proposed algorithm has been verified on an ALMR for the power distribution network and archived good results.

Automated summary

With the development and application of autonomous live-line maintenance robots (ALMRs), uneven illumination in outdoor environments poses challenges for visual feedback and target recognition. This study proposes an image enhancement method based on image brightness segmentation and multi-methods fusion to improve the visual performance of ALMRs. The experimental results show that the proposed method achieves good results on multiple indices and has a relatively short processing time.