用于检测和分割输电塔和输电线的航空图像数据集(TTPLA)

从航空图像中准确检测和分割输电塔（TT）和输电线（PL）对保护电网安全和低空无人机安全起着关键作用。同时，输电塔和输电线的航空图像也给从事物体检测和分割工作的计算机视觉研究人员带来了许多新的挑战--输电线又细又长，可能会呈现出与背景相似的颜色；输电塔的形状多种多样，很可能是由稀疏程度不同的线路结构组成；不同图像的背景场景、光照和物体大小都会有很大差异。本文收集并发布了一个新的 TT/PL 空中图像（TTPLA）数据集，该数据集由分辨率为 3,840 2,160 像素的 1,100 幅图像以及人工标注的 8,987 个 TT 和 PL 实例组成。Accurately detecting and segmenting Transmission Towers (TT) and Power Lines (PL) from aerial images plays a critical role in safeguarding both power grid security and the safety of low-altitude unmanned aerial vehicles (UAVs). Meanwhile, aerial images of TT and PL pose numerous novel challenges for computer vision researchers engaged in object detection and segmentation tasks: Power lines are thin and elongated, and may exhibit colors similar to their background; transmission towers have diverse shapes, often composed of line structures with varying sparsity; additionally, background scenes, lighting conditions and object sizes vary significantly across different images. This paper collects and releases a novel aerial image dataset for TT and PL (named TTPLA), which consists of 1,100 images with a resolution of 3,840 × 2,160 pixels, along with 8,987 manually annotated TT and PL instances.

搜集汇总

背景与挑战

背景概述

TTPLA是一个航空图像数据集，专门用于检测和分割输电塔和输电线，以支持电网安全和低空无人机安全应用。该数据集包含1,100张高分辨率图像（3,840×2,160像素）和8,987个人工标注实例，挑战在于输电线细长、颜色与背景相似，输电塔形状多样且结构稀疏，图像背景和光照差异大。