Measurement of buried depth and inclination rate of concrete poles based on binocular vision

Abstract

Aiming at the low efficiency of the traditional detection methods of buried depth and inclination rate of concrete poles, a method of measuring concrete poles based on binocular vision was proposed. Firstly, an improved DeeplabV3+ semantic segmentation algorithm is proposed. Based on the original model structure, the backbone feature extraction network is modified, the feature fusion method is optimized, and the improved CBAM attention mechanism is added to reduce the model complexity and improved the accuracy of concrete pole area segmentation. Secondly, the sub-pixel edge extraction algorithm based on local area effect is used to determine the edge of the concrete pole in the image segmentation area, and the least squares method is used to fit the edge to determine the precise feature points. Finally, the coordinate transformation of binocular vision is used to calculate the depth and inclination of the concrete pole. Experiments show that the method has a buried depth measurement error of less than 10 cm, an error rate of less than 5%, and an inclination measurement error of less than 0.3°, which provides an automated concrete pole buried depth and inclination rate measurement solution