露天矿区场景复杂,行车障碍物检测受扬尘和颗粒物等粉尘噪声干扰严重,难以准确识别障碍物,尤其是光线较差的夜间,不利于做出正确决策,从而影响无人作业的安全性和整体效率。针对以上问题,提出了一种基于YOLOv8n模型的露天矿区行车障碍...露天矿区场景复杂,行车障碍物检测受扬尘和颗粒物等粉尘噪声干扰严重,难以准确识别障碍物,尤其是光线较差的夜间,不利于做出正确决策,从而影响无人作业的安全性和整体效率。针对以上问题,提出了一种基于YOLOv8n模型的露天矿区行车障碍物检测算法YOLOv8n-Enhanced。该算法主要从3个方面进行了改进,具体包括:首先,针对受粉尘噪声干扰严重和夜间光线不足的问题,提出了C2fCA模块结构,提高了模型特征提取能力;其次,使用轻量级卷积技术GSConv和VoV-GSCSP模块,减轻模型复杂性,实现检测器更高的计算成本效益;最后,使用WIOU损失函数,提高了模型泛化能力。试验结果表明:改进算法在保持实时性的前提下,可将YOLOv8n的平均精度(mean Average Precision,mAP)分别提高1.8%和2.6%,实现白天与夜间场景下不同尺度的障碍物识别。展开更多
街道场景视频实例分割是无人驾驶技术研究中的关键问题之一,可为车辆在街道场景下的环境感知和路径规划提供决策依据.针对现有方法存在多纵横比锚框应用单一感受野采样导致边缘特征提取不充分以及高层特征金字塔空间细节位置信息匮乏的...街道场景视频实例分割是无人驾驶技术研究中的关键问题之一,可为车辆在街道场景下的环境感知和路径规划提供决策依据.针对现有方法存在多纵横比锚框应用单一感受野采样导致边缘特征提取不充分以及高层特征金字塔空间细节位置信息匮乏的问题,本文提出锚框校准和空间位置信息补偿视频实例分割(Anchor frame calibration and Spatial position information compensation for Video Instance Segmentation,AS-VIS)网络.首先,在预测头3个分支中添加锚框校准模块实现同锚框纵横比匹配的多类型感受野采样,解决目标边缘提取不充分问题.其次,设计多感受野下采样模块将各种感受野采样后的特征融合,解决下采样信息缺失问题.最后,应用多感受野下采样模块将特征金字塔低层目标区域激活特征映射嵌入到高层中实现空间位置信息补偿,解决高层特征空间细节位置信息匮乏问题.在Youtube-VIS标准库中提取街道场景视频数据集,其中包括训练集329个视频和验证集53个视频.实验结果与YolactEdge检测和分割精度指标定量对比表明,锚框校准平均精度分别提升8.63%和5.09%,空间位置信息补偿特征金字塔平均精度分别提升7.76%和4.75%,AS-VIS总体平均精度分别提升9.26%和6.46%.本文方法实现了街道场景视频序列实例级同步检测、跟踪与分割,为无人驾驶车辆环境感知提供有效的理论依据.展开更多
As hybrid vehicles introduced the motor, the vehicle structure has a significant change in the power matching. A driver-vehicle-road closed-loop semi-physical simulation system, which makes real driving parts together...As hybrid vehicles introduced the motor, the vehicle structure has a significant change in the power matching. A driver-vehicle-road closed-loop semi-physical simulation system, which makes real driving parts together with the simulation car, will bring convenience to the new car design. We used the computer software to simulate the road with a slope, curve and some other features based on the actual road condition, and analyzed the whole road scene in addition to geometry and physical characteristics. Analyzing and constructing the vehicle dynamics basic template, appropriate changes to the template can obtain the desired vehicle dynamics model with an external device to control the model vehicle. It combined the physical operation system with visual display, which gave us real driving feelings and increased the vehicle design predictive accuracy.展开更多
文摘露天矿区场景复杂,行车障碍物检测受扬尘和颗粒物等粉尘噪声干扰严重,难以准确识别障碍物,尤其是光线较差的夜间,不利于做出正确决策,从而影响无人作业的安全性和整体效率。针对以上问题,提出了一种基于YOLOv8n模型的露天矿区行车障碍物检测算法YOLOv8n-Enhanced。该算法主要从3个方面进行了改进,具体包括:首先,针对受粉尘噪声干扰严重和夜间光线不足的问题,提出了C2fCA模块结构,提高了模型特征提取能力;其次,使用轻量级卷积技术GSConv和VoV-GSCSP模块,减轻模型复杂性,实现检测器更高的计算成本效益;最后,使用WIOU损失函数,提高了模型泛化能力。试验结果表明:改进算法在保持实时性的前提下,可将YOLOv8n的平均精度(mean Average Precision,mAP)分别提高1.8%和2.6%,实现白天与夜间场景下不同尺度的障碍物识别。
文摘街道场景视频实例分割是无人驾驶技术研究中的关键问题之一,可为车辆在街道场景下的环境感知和路径规划提供决策依据.针对现有方法存在多纵横比锚框应用单一感受野采样导致边缘特征提取不充分以及高层特征金字塔空间细节位置信息匮乏的问题,本文提出锚框校准和空间位置信息补偿视频实例分割(Anchor frame calibration and Spatial position information compensation for Video Instance Segmentation,AS-VIS)网络.首先,在预测头3个分支中添加锚框校准模块实现同锚框纵横比匹配的多类型感受野采样,解决目标边缘提取不充分问题.其次,设计多感受野下采样模块将各种感受野采样后的特征融合,解决下采样信息缺失问题.最后,应用多感受野下采样模块将特征金字塔低层目标区域激活特征映射嵌入到高层中实现空间位置信息补偿,解决高层特征空间细节位置信息匮乏问题.在Youtube-VIS标准库中提取街道场景视频数据集,其中包括训练集329个视频和验证集53个视频.实验结果与YolactEdge检测和分割精度指标定量对比表明,锚框校准平均精度分别提升8.63%和5.09%,空间位置信息补偿特征金字塔平均精度分别提升7.76%和4.75%,AS-VIS总体平均精度分别提升9.26%和6.46%.本文方法实现了街道场景视频序列实例级同步检测、跟踪与分割,为无人驾驶车辆环境感知提供有效的理论依据.
基金Funded by the National Natural Science Foundation of China(No.51305475)Chongqing Research Program of Basic Research and Frontier Technology(No.cstc2013jcyj A60004)the Scientific and Technological Research Program of Chongqing Municipal Education Commission(No.KJ1500927)
文摘As hybrid vehicles introduced the motor, the vehicle structure has a significant change in the power matching. A driver-vehicle-road closed-loop semi-physical simulation system, which makes real driving parts together with the simulation car, will bring convenience to the new car design. We used the computer software to simulate the road with a slope, curve and some other features based on the actual road condition, and analyzed the whole road scene in addition to geometry and physical characteristics. Analyzing and constructing the vehicle dynamics basic template, appropriate changes to the template can obtain the desired vehicle dynamics model with an external device to control the model vehicle. It combined the physical operation system with visual display, which gave us real driving feelings and increased the vehicle design predictive accuracy.