Overtaking is a crucial maneuver in road transportation that requires a clear view of the road ahead.However,limited visibility of ahead vehicles can often make it challenging for drivers to assess the safety of overt...Overtaking is a crucial maneuver in road transportation that requires a clear view of the road ahead.However,limited visibility of ahead vehicles can often make it challenging for drivers to assess the safety of overtaking maneuvers,leading to accidents and fatalities.In this paper,we consider atrous convolution,a powerful tool for explicitly adjusting the field-of-view of a filter as well as controlling the resolution of feature responses generated by Deep Convolutional Neural Networks in the context of semantic image segmentation.This article explores the potential of seeing-through vehicles as a solution to enhance overtaking safety.See-through vehicles leverage advanced technologies such as cameras,sensors,and displays to provide drivers with a real-time view of the vehicle ahead,including the areas hidden from their direct line of sight.To address the problems of safe passing and occlusion by huge vehicles,we designed a see-through vehicle system in this study,we employed a windshield display in the back car together with cameras in both cars.The server within the back car was used to segment the car,and the segmented portion of the car displayed the video from the front car.Our see-through system improves the driver’s field of vision and helps him change lanes,cross a large car that is blocking their view,and safely overtake other vehicles.Our network was trained and tested on the Cityscape dataset using semantic segmentation.This transparent technique will instruct the driver on the concealed traffic situation that the front vehicle has obscured.For our findings,we have achieved 97.1% F1-score.The article also discusses the challenges and opportunities of implementing see-through vehicles in real-world scenarios,including technical,regulatory,and user acceptance factors.展开更多
[目的/意义]针对小麦叶片病虫害在自然环境下形态和颜色特征较为复杂、区分度较低等特点,提出一种高质量高效的病虫害检测模型,即YOLOv8-SS (You Only Look Once Version 8-SS),为病虫害的预防与科学化治理提供准确的依据。[方法]基于YO...[目的/意义]针对小麦叶片病虫害在自然环境下形态和颜色特征较为复杂、区分度较低等特点,提出一种高质量高效的病虫害检测模型,即YOLOv8-SS (You Only Look Once Version 8-SS),为病虫害的预防与科学化治理提供准确的依据。[方法]基于YOLOv8算法,采用改进的轻量级卷积神经网络ShuffleNet V2作为主干网络提取图像特征即YOLOv8-S,在保持检测精度的同时,减少模型的参数数量和计算负载;在此基础上增加小目标检测层和注意力机制SEnet (Squeeze and Excitation Network),对YOLOv8-S进行改进,在不降低检测速度和不损失模型轻量化程度的情况下提高检测精度,提出YOLOv8-SS小麦叶片病虫害检测模型。[结果与讨论]YOLOv8-SS模型在实验数据集上的平均识别精度和检测准确率分别达89.41%和91.00%,对比原模型分别提高10.11%和7.42%。因此,本研究所提出的方法可显著提高农作物病虫害的检测鲁棒性,并增强模型对小目标图像特征的提取能力,从而高效准确地进行病虫害的检测和识别。[结论]本研究使用的方法具有广泛适用性,可应用于大规模农作物病虫害检测的实际场景中。展开更多
基金financially supported by the Ministry of Trade,Industry and Energy(MOTIE)and Korea Institute for Advancement of Technology(KIAT)through the International Cooperative R&D Program(Project No.P0016038)supported by the MSIT(Ministry of Sci-ence and ICT),Korea,under the ITRC(Information Technology Research Center)support program(IITP-2022-RS-2022-00156354)supervised by the IITP(Institute for Information&Communications Technology Planning&Evaluation).
文摘Overtaking is a crucial maneuver in road transportation that requires a clear view of the road ahead.However,limited visibility of ahead vehicles can often make it challenging for drivers to assess the safety of overtaking maneuvers,leading to accidents and fatalities.In this paper,we consider atrous convolution,a powerful tool for explicitly adjusting the field-of-view of a filter as well as controlling the resolution of feature responses generated by Deep Convolutional Neural Networks in the context of semantic image segmentation.This article explores the potential of seeing-through vehicles as a solution to enhance overtaking safety.See-through vehicles leverage advanced technologies such as cameras,sensors,and displays to provide drivers with a real-time view of the vehicle ahead,including the areas hidden from their direct line of sight.To address the problems of safe passing and occlusion by huge vehicles,we designed a see-through vehicle system in this study,we employed a windshield display in the back car together with cameras in both cars.The server within the back car was used to segment the car,and the segmented portion of the car displayed the video from the front car.Our see-through system improves the driver’s field of vision and helps him change lanes,cross a large car that is blocking their view,and safely overtake other vehicles.Our network was trained and tested on the Cityscape dataset using semantic segmentation.This transparent technique will instruct the driver on the concealed traffic situation that the front vehicle has obscured.For our findings,we have achieved 97.1% F1-score.The article also discusses the challenges and opportunities of implementing see-through vehicles in real-world scenarios,including technical,regulatory,and user acceptance factors.
文摘[目的/意义]针对小麦叶片病虫害在自然环境下形态和颜色特征较为复杂、区分度较低等特点,提出一种高质量高效的病虫害检测模型,即YOLOv8-SS (You Only Look Once Version 8-SS),为病虫害的预防与科学化治理提供准确的依据。[方法]基于YOLOv8算法,采用改进的轻量级卷积神经网络ShuffleNet V2作为主干网络提取图像特征即YOLOv8-S,在保持检测精度的同时,减少模型的参数数量和计算负载;在此基础上增加小目标检测层和注意力机制SEnet (Squeeze and Excitation Network),对YOLOv8-S进行改进,在不降低检测速度和不损失模型轻量化程度的情况下提高检测精度,提出YOLOv8-SS小麦叶片病虫害检测模型。[结果与讨论]YOLOv8-SS模型在实验数据集上的平均识别精度和检测准确率分别达89.41%和91.00%,对比原模型分别提高10.11%和7.42%。因此,本研究所提出的方法可显著提高农作物病虫害的检测鲁棒性,并增强模型对小目标图像特征的提取能力,从而高效准确地进行病虫害的检测和识别。[结论]本研究使用的方法具有广泛适用性,可应用于大规模农作物病虫害检测的实际场景中。
