Visual distance measurement system for overhead transmission lines based on improved SGBM algorithm

Overhead transmission lines serve as the primary channels for power transmission and are essential to residents’ daily lives and industrial production, spanning major roads across cities, districts, counties, and towns nationwide. Municipal engineering construction is a key initiative for promoting urban development and improving the quality of life, involving numerous large-scale projects that frequently employ heavy construction machinery. To prevent external damage to overhead power transmission lines and ensure the operational safety of construction machinery and workers, developing a binocular vision ranging-based safety warning system for overhead transmission lines targeting construction machinery is essential.

The hardware architecture of the developed safety warning system primarily consists of a perception layer, a computation layer, and an application layer. The perception layer includes CMOS binocular cameras for real-time cable image acquisition and Raspberry Pi modules for low-latency local area network (LAN) image transmission. The computation layer incorporates a host computer for efficient image processing, and the application layer includes voice prompt chips, an electromagnetic buzzer, LED indicator beads, and a power supply module. Key software modules were also carefully selected and refined to enhance system performance under realistic operating conditions. Zhang’s calibration method, widely recognized for its reliability, was adopted for three-dimensional (3D) calibration, and the Bouguet rectification method was employed for epipolar rectification and image distortion correction. An improved semi-global block matching (SGBM) algorithm was proposed to enhance stereo matching performance. Adaptive windows, noise tolerance variables, and RGB channel absolute color difference (ACD) costs were simultaneously incorporated into the conventional Census Transform (CT) method to yield an improved cost calculation approach. A four-path cost aggregation strategy was then applied for initial matching cost aggregation, followed by the Winner-Take-All (WTA) algorithm to determine the initial disparity of the target image. Subsequently, the initial disparity was optimized using a median filtering algorithm to obtain the optimal disparity.

Compared with the traditional SGBM algorithm, the image processing time of the improved SGBM algorithm (3.528 ms) increased by 18.26%; however, its overall error rate (2.79%) and non-occlusion error rate (2.16%) decreased by 95.08% and 95.04%, respectively. The relative error between the measured values from the binocular vision-based ranging system and the actual distances consistently remained within the range of 0.691% to 2.482%.

The binocular vision-based overhead transmission line ranging system developed in this study demonstrates high measurement accuracy, high detection efficiency, and reliable operational stability under realistic conditions, effectively ensuring the safety of construction machinery operations and preventing external damage to overhead power transmission lines.