| 摘要: |
| 电力线是一类形状细长、特征稀疏、随着视角的变化容易混淆在大量背景信息中的特殊障碍物，常规电力线检测识别算法得到的目标框对电力线所在位置的估计不够准确。为此，提出了一种相对角度估计方法，基于常规电力线目标检测与识别算法，并结合电力线相对角度估计，从而提高电力线的检测识别过程中所在位置的精度。相比电力线绝对角度回归的方法，提出的相对角度估计方法容易训练易收敛，计算量小，适用于实时性要求较高的应用场合。 |
| 关键词: 直升机飞行安全 目标检测识别 电力线检测 深度学习 |
| DOI： |
|
| 基金项目: |
|
| A power line detection method for helicopter obstacle avoidance |
| ZHU Peipei |
| (Southwest China Institute of Electronic Technology,Chengdu 610036，China) |
| Abstract: |
| Power line is a special,linear and sparsely characterized obstacle.It is hard to accurately estimate the location of power line,which is easily getting lost in the background.To improve the accuracy of power lines location,the conventional object detection model is combined with relative angle calculation.Compared with calculation of absolute angles of power lines,the proposed method with less computation is easier to train and converge.It can be applied in the situation requiring high realtime. |
| Key words: helicopter flight safty object detection and identification power line detection deep learning |
《电讯技术》编辑部 