摘要
针对无人机避障系统中视觉防撞的关键技术进行研究,改进YOLOv4-tiny目标检测算法,并提出一种基于目标检测的视觉防撞方法。结合MobileNet的可分离卷积结构,优化YOLOv4-tiny目标检测网络,以此提高检测精度。通过去除非检测目标区域的冗余信息,提高SURF匹配算法的效率,由最小二乘法对目标区域进行三维解算,将解算的深度与无人机的位姿信息作为防撞判定区域视野尺度的依据,并结合目标检测的像素位置,实现无人机视觉防撞。最后,利用标准数据集对改进后的网络模型进行分析,mAP值为69%,并搭建无人机视觉通讯模块进行视觉防撞测试,最终验证了方法的合理性。
Aiming at the key technologies of visual collision avoidance in UAV obstacle avoidance systemthe YOLOv4-tiny target detection algorithm is improvedand a visual collision avoidance method based on target detection is proposed.With the separable convolution structure of MobileNetthe YOLOv4-tiny target detection network is optimized to improve the detection accuracy.The efficiency of the SURF matching algorithm is improved by removing the redundant information of the undetected target area.The target area is calculated in three dimensions by least squares methodand the calculated depth and pose information of UAV are treated as the basis of visual field scale of the collision avoidance determination areathe visual collision avoidance of UAV is realized through the integration of the pixel position of target detection.Finallythe improved network model is analyzed by using standard data setthe mAP value is 69%and the UAV visual communication module is built for visual collision avoidance testwhich verifies the rationality of the method.
作者
成怡
郑腾龙
CHENG Yi;ZHENG Tenglong(Tianjin Polytechnic University ,Tianjin 300000 China)
出处
《电光与控制》
CSCD
北大核心
2021年第10期31-35,共5页
Electronics Optics & Control
基金
天津市自然科学基金(18JCYBJC88300,18JCYBJC88400)。
关键词
双目视觉
视觉避障
深度学习
目标检测
特征匹配
binocular vision
visual obstacle avoidance
deep learning
target detection
feature matching