An investigation on the ventral diverterless high offset S-shaped inlet is carried out at Mach numbers from 0.600 to 1.534, angles of attack from -4° to 9.4°, and yaw angles from 0° to 8°. Results ...An investigation on the ventral diverterless high offset S-shaped inlet is carried out at Mach numbers from 0.600 to 1.534, angles of attack from -4° to 9.4°, and yaw angles from 0° to 8°. Results indicate： （1） a large region of low total pressure exists at the lower part of the inlet exit caused by the counter-rotating vortices in the S-shaped duct; （2） the performances of the inlet at Mach number 1.000 reach almost the highest, so the propulsion system could work efficiently in terms of aerodynamics; （3） the total pressure recovery increases slowly at first and then remains unvaried as the Mach number rises from 0.6 to 1.0, however, it does in an opposite manner in the conventional diverter-equipped S-shaped inlet; （4） the performances of the inlet are generally insensitive to angles of attack from -4° to 9.4° and yaw angles from 0° to 8° at Mach number 0.850, and angles of attack from -2° to 6° and yaw angles from 0° to 5° at Mach number 1.534.展开更多
针对传统中药材检测任务中识别效率低、受主观因素影响较大的问题,文章选取77种中药材作为研究对象。采用自行拍摄图像和在互联网获取图像的方式,并结合旋转平移、高斯噪声等数据增强技术,最终构建了一个包含4万多张图像的数据集。在模...针对传统中药材检测任务中识别效率低、受主观因素影响较大的问题,文章选取77种中药材作为研究对象。采用自行拍摄图像和在互联网获取图像的方式,并结合旋转平移、高斯噪声等数据增强技术,最终构建了一个包含4万多张图像的数据集。在模型改进方面,对第八代只看一次目标检测算法(You Only Look Once version 8,YOLOv8)的Backbone部分进行了针对性的优化,引入了DSConv和Biformer注意力机制。DSConv能够自适应地关注细长和曲折的局部特征,而Biformer则通过双层路由机制,实现了内容感知的稀疏模式,提高了模型对图像细节和关键目标的识别能力。实验结果表明,改进后的YOLOv8模型的精确率、召回率和平均精度分别达到了96.4%、98.0%和97.7%,相较于原模型的精确率和平均精度分别增长了1.7百分点和1.0百分点。在中药材检测任务上取得了显著的性能提升效果。展开更多
基金National Basic Research Program of China (5130802)
文摘An investigation on the ventral diverterless high offset S-shaped inlet is carried out at Mach numbers from 0.600 to 1.534, angles of attack from -4° to 9.4°, and yaw angles from 0° to 8°. Results indicate： （1） a large region of low total pressure exists at the lower part of the inlet exit caused by the counter-rotating vortices in the S-shaped duct; （2） the performances of the inlet at Mach number 1.000 reach almost the highest, so the propulsion system could work efficiently in terms of aerodynamics; （3） the total pressure recovery increases slowly at first and then remains unvaried as the Mach number rises from 0.6 to 1.0, however, it does in an opposite manner in the conventional diverter-equipped S-shaped inlet; （4） the performances of the inlet are generally insensitive to angles of attack from -4° to 9.4° and yaw angles from 0° to 8° at Mach number 0.850, and angles of attack from -2° to 6° and yaw angles from 0° to 5° at Mach number 1.534.
文摘针对传统中药材检测任务中识别效率低、受主观因素影响较大的问题,文章选取77种中药材作为研究对象。采用自行拍摄图像和在互联网获取图像的方式,并结合旋转平移、高斯噪声等数据增强技术,最终构建了一个包含4万多张图像的数据集。在模型改进方面,对第八代只看一次目标检测算法(You Only Look Once version 8,YOLOv8)的Backbone部分进行了针对性的优化,引入了DSConv和Biformer注意力机制。DSConv能够自适应地关注细长和曲折的局部特征,而Biformer则通过双层路由机制,实现了内容感知的稀疏模式,提高了模型对图像细节和关键目标的识别能力。实验结果表明,改进后的YOLOv8模型的精确率、召回率和平均精度分别达到了96.4%、98.0%和97.7%,相较于原模型的精确率和平均精度分别增长了1.7百分点和1.0百分点。在中药材检测任务上取得了显著的性能提升效果。
