| 基于高斯模型和YOLOv3的光伏电场巡检无人机避障目标检测方法 |
 点此下载全文 |
| 引用本文：唐明1，张宇宁1，李夏叶2，温贤茂1，刘诗剑3.基于高斯模型和YOLOv3的光伏电场巡检无人机避障目标检测方法[J].计算技术与自动化,2025,(1):141-146 |
| 摘要点击次数: 65 |
| 全文下载次数: 40 |
|
|
| 中文摘要:目前常规的无人机避障目标检测方法主要采用超声波技术实现对目标轮廓的获取，由于缺乏对图像的形态学处理，导致检测精度较差。对此，提出了基于高斯模型和YOLOv3的光伏电场巡检无人机避障目标检测方法。首先结合混合高斯模型，对巡检图像的前景与背景进行分离处理，并采用滤波算法对前景图像进行去噪处理。然后结合分类损失函数以及位置损失函数，对特征点的梯度值以及方向进行计算，实现障碍物图像的特征提取，最后对提取到的避障目标边缘进行补偿处理，实现避障目标检测。在实验中，对提出的方法进行了避障目标检测精度的检验。最终的测试结果表明，采用提出的方法进行无人机避障检测时，算法的mAP值较高，具备较为理想的检测精度。 |
| 中文关键词:高斯模型 光伏电场 无人机巡检 检测方法 |
| |
| Obstacle Avoidance Target Detection Method for Photovoltaic Power Field Inspection UAV Based on Gaussian Model and YOLOv3 |
|
|
| Abstract:Current conventional UAV obstacle avoidance target detection methods mainly use ultrasonic technology to realize the acquisition of the target contour, which results in poor detection accuracy due to the lack of morphological processing of the image. In this regard, an obstacle avoidance target detection method for photovoltaic power field inspection UAV based on Gaussian model and YOLOv3 is proposed. Firstly, the foreground and background of the inspection image are separated by combining the hybrid Gaussian model, and the foreground image is denoised by using a filtering algorithm. Then the gradient value and direction of the feature points are calculated by combining the classification loss function and the position loss function to realize the feature extraction of the obstacle image, and finally the extracted edge of the obstacle avoidance target is compensated to realize the detection of the obstacle avoidance target. In the experiments, the accuracy of obstacle avoidance target detection is tested for the proposed method. The final test results show that when the proposed method is used for UAV obstacle avoidance detection, the algorithm has a high mAP value and has a more ideal detection accuracy. |
| keywords:Gaussian model photovoltaic (PV) electric field UAV inspection detection methods |
| 查看全文 查看/发表评论 下载pdf阅读器 |
次