近年来,由于基于图像识别的粉尘检测方法不存在安装和检测范围局限性等问题,因此得到了充分重视和发展,但现有方法实时性和准确性仍需提升。为此,提出了一种基于改进YOLOv5算法的粉尘图像检测方法。首先,对现有YOLOv5算法主干网络以及N...近年来,由于基于图像识别的粉尘检测方法不存在安装和检测范围局限性等问题,因此得到了充分重视和发展,但现有方法实时性和准确性仍需提升。为此,提出了一种基于改进YOLOv5算法的粉尘图像检测方法。首先,对现有YOLOv5算法主干网络以及Neck网络进行改进,将轻量化网络GhostNet替换原有主干网络,以降低网络参数,再输出3个特征层;然后,针对主干网络输出的3个特征层,施加注意力机制CA,增加网络精度;最后,设计消融实验和对比实验验证改进算法的有效性。结果表明:改进算法的平均检测精度mAP(mean Average Precision)能达到92.11%,检测速度达37帧/s。展开更多
The fundamental access method of IEEE 802.11 is a DCF known as carrier sense multiple access with collision avoidance (CSMA/CA) scheme with exponential back-off. RTS__threshold is used to determine whether to deploy R...The fundamental access method of IEEE 802.11 is a DCF known as carrier sense multiple access with collision avoidance (CSMA/CA) scheme with exponential back-off. RTS__threshold is used to determine whether to deploy RTS/CTS access method. This threshold should vary with the number of contending stations which contend wireless media to get better throughput. The paper proposes an algorithm which estimates the number of contending stations in BSS. The algorithm is shown to be accurate which is verified by elaborate simulations.展开更多
摘要近年来,由于基于图像识别的粉尘检测方法不存在安装和检测范围局限性等问题,因此得到了充分重视和发展,但现有方法实时性和准确性仍需提升。为此,提出了一种基于改进YOLOv5算法的粉尘图像检测方法。首先,对现有YOLOv5算法主干网络以及Neck网络进行改进,将轻量化网络GhostNet替换原有主干网络,以降低网络参数,再输出3个特征层;然后,针对主干网络输出的3个特征层,施加注意力机制CA,增加网络精度;最后,设计消融实验和对比实验验证改进算法的有效性。结果表明:改进算法的平均检测精度mAP(mean Average Precision)能达到92.11%,检测速度达37帧/s。
摘要The fundamental access method of IEEE 802.11 is a DCF known as carrier sense multiple access with collision avoidance (CSMA/CA) scheme with exponential back-off. RTS__threshold is used to determine whether to deploy RTS/CTS access method. This threshold should vary with the number of contending stations which contend wireless media to get better throughput. The paper proposes an algorithm which estimates the number of contending stations in BSS. The algorithm is shown to be accurate which is verified by elaborate simulations.