Objective: To test the hypothesis that the N10 far field potential in median nerve somatosensory evoked potentials is generated by the motor axons by examini ng patients with amyotrophic lateral sclerosis (ALS). Metho...Objective: To test the hypothesis that the N10 far field potential in median nerve somatosensory evoked potentials is generated by the motor axons by examini ng patients with amyotrophic lateral sclerosis (ALS). Methods: Subjects were 5 A LS patients showing pronounced or complete denervation of median inner vated s mall hand muscles. We evaluated N10 over scalp, and proximal plexus volleys (PPV s) at lateral or anterior cervical electrode. Results: N10 and PPVs were definit ely preserved for every ALS subject. N10 amplitudes of ALS subjects were even si gnificantly larger than control subjects. In one ALS patient completely lacking motor axons, N10 was larger than the largest one among control subjects. Conclus ions: Present results clearly indicate that N10 is not predominantly generated b y motor axons but by the whole median nerve dominated by sensory axons. We propo se a theory that N10 is a junctional potential generated by the entrance of the median nerve into bone at the intervertebral foramen, producing a positive pole at the non cephalic reference electrode. Significantly larger N10 in ALS subjec ts may be due to the lack of cancellation by slower motor axons. Significance: T he hypothesis that N10 is generated by motor axons is refuted, and a new theory of its generation is presented.展开更多
针对玉米杂草识别过程中因光照变化导致识别精确度低及漏检问题,该研究以幼苗期玉米及其伴生杂草为研究对象,设计一种基于WEED-YOLOv10的玉米杂草检测方法。首先,通过无人机快速采集田间高分辨率图像构建了玉米杂草数据集;其次,以YOLOv...针对玉米杂草识别过程中因光照变化导致识别精确度低及漏检问题,该研究以幼苗期玉米及其伴生杂草为研究对象,设计一种基于WEED-YOLOv10的玉米杂草检测方法。首先,通过无人机快速采集田间高分辨率图像构建了玉米杂草数据集;其次,以YOLOv10n为基线网络,将骨干网络替换为ConvNeXtV2以增强特征提取能力;继而,为避免因模块拼接可能带来的信息冗余或丢失问题提升对光照干扰的鲁棒性,嵌入CBAM注意力机制;然后,引入SlimNeck结构优化网络计算效率,有效平衡了模型计算资源消耗与特征表征能力;最后,使用Focaler-EIoU损失函数进一步提高模型定位精度。试验结果表明,WEED-YOLOv10在精确率、召回率、mAP@50、mAP@50:95和F1分数上分别达到85.4%、88.1%、90.9%、48.5%和86.7%,较基准模型分别提升了2.4、2.9、3.5、7.0、2.6个百分点,各项精度指标均优于其他对比模型,部署在NVIDIA Jetson orin NX上的图片推理速度达到28.7帧/s,实现了检测速度与精度的平衡。进一步地,基于WEED-YOLOv10开发对靶喷药系统,该系统实时捕捉并解析来自模型的识别信号,实现对除草喷施装置的精准调控。田间试验结果显示,对靶喷药系统施药准确率为93.7%,喷洒覆盖率为90.5%,对靶偏差为1.45cm,杂草实时检测速度为20.1帧/s,实现了自动化的玉米田间除草作业。该研究为复杂光照场景下农田杂草治理提供了可靠的技术方案,对推动农业智能化作业具有重要意义。展开更多
为解决现有道路裂缝检测模型精度低、易漏检误检等问题,提出了一种基于混合局部通道注意力机制的YOLOv10n道路裂缝检测模型。针对道路裂缝细长且不规则的形状特点,首先在模型特征提取网络引入混合局部通道注意力机制(mixed local channe...为解决现有道路裂缝检测模型精度低、易漏检误检等问题,提出了一种基于混合局部通道注意力机制的YOLOv10n道路裂缝检测模型。针对道路裂缝细长且不规则的形状特点,首先在模型特征提取网络引入混合局部通道注意力机制(mixed local channel attention,MLCA),增强模型对道路裂缝特征的提取能力;其次,采用重参数化泛化特征金字塔网络(reparameterized generalized feature pyramid network,RepGFPN)优化原始颈部网络,充分融合多尺度下的裂缝特征信息;最后使用Focaler-IoU替换CIoU损失函数,调整模型训练不同裂缝样本的权重,加快收敛速度。在RDD2022_China数据集上的实验结果表明,改进后的模型相较于原始YOLOv10n模型检测准确率提升4.4%,平均精度均值(mean average precision,mAP)提高2.9%。与其他主流目标检测模型相比,改进后的模型在准确率、召回率和计算成本等方面均展现出最佳性能,验证了本文方法在道路裂缝检测任务中的有效性和优越性。展开更多
文摘Objective: To test the hypothesis that the N10 far field potential in median nerve somatosensory evoked potentials is generated by the motor axons by examini ng patients with amyotrophic lateral sclerosis (ALS). Methods: Subjects were 5 A LS patients showing pronounced or complete denervation of median inner vated s mall hand muscles. We evaluated N10 over scalp, and proximal plexus volleys (PPV s) at lateral or anterior cervical electrode. Results: N10 and PPVs were definit ely preserved for every ALS subject. N10 amplitudes of ALS subjects were even si gnificantly larger than control subjects. In one ALS patient completely lacking motor axons, N10 was larger than the largest one among control subjects. Conclus ions: Present results clearly indicate that N10 is not predominantly generated b y motor axons but by the whole median nerve dominated by sensory axons. We propo se a theory that N10 is a junctional potential generated by the entrance of the median nerve into bone at the intervertebral foramen, producing a positive pole at the non cephalic reference electrode. Significantly larger N10 in ALS subjec ts may be due to the lack of cancellation by slower motor axons. Significance: T he hypothesis that N10 is generated by motor axons is refuted, and a new theory of its generation is presented.
文摘针对玉米杂草识别过程中因光照变化导致识别精确度低及漏检问题,该研究以幼苗期玉米及其伴生杂草为研究对象,设计一种基于WEED-YOLOv10的玉米杂草检测方法。首先,通过无人机快速采集田间高分辨率图像构建了玉米杂草数据集;其次,以YOLOv10n为基线网络,将骨干网络替换为ConvNeXtV2以增强特征提取能力;继而,为避免因模块拼接可能带来的信息冗余或丢失问题提升对光照干扰的鲁棒性,嵌入CBAM注意力机制;然后,引入SlimNeck结构优化网络计算效率,有效平衡了模型计算资源消耗与特征表征能力;最后,使用Focaler-EIoU损失函数进一步提高模型定位精度。试验结果表明,WEED-YOLOv10在精确率、召回率、mAP@50、mAP@50:95和F1分数上分别达到85.4%、88.1%、90.9%、48.5%和86.7%,较基准模型分别提升了2.4、2.9、3.5、7.0、2.6个百分点,各项精度指标均优于其他对比模型,部署在NVIDIA Jetson orin NX上的图片推理速度达到28.7帧/s,实现了检测速度与精度的平衡。进一步地,基于WEED-YOLOv10开发对靶喷药系统,该系统实时捕捉并解析来自模型的识别信号,实现对除草喷施装置的精准调控。田间试验结果显示,对靶喷药系统施药准确率为93.7%,喷洒覆盖率为90.5%,对靶偏差为1.45cm,杂草实时检测速度为20.1帧/s,实现了自动化的玉米田间除草作业。该研究为复杂光照场景下农田杂草治理提供了可靠的技术方案,对推动农业智能化作业具有重要意义。
文摘为解决现有道路裂缝检测模型精度低、易漏检误检等问题,提出了一种基于混合局部通道注意力机制的YOLOv10n道路裂缝检测模型。针对道路裂缝细长且不规则的形状特点,首先在模型特征提取网络引入混合局部通道注意力机制(mixed local channel attention,MLCA),增强模型对道路裂缝特征的提取能力;其次,采用重参数化泛化特征金字塔网络(reparameterized generalized feature pyramid network,RepGFPN)优化原始颈部网络,充分融合多尺度下的裂缝特征信息;最后使用Focaler-IoU替换CIoU损失函数,调整模型训练不同裂缝样本的权重,加快收敛速度。在RDD2022_China数据集上的实验结果表明,改进后的模型相较于原始YOLOv10n模型检测准确率提升4.4%,平均精度均值(mean average precision,mAP)提高2.9%。与其他主流目标检测模型相比,改进后的模型在准确率、召回率和计算成本等方面均展现出最佳性能,验证了本文方法在道路裂缝检测任务中的有效性和优越性。
