棉田虫害的快速检测与准确识别是预防棉田虫害、提高棉花品质的重要前提。针对真实棉田环境下昆虫相似度高、背景干扰严重的问题,该研究提出一种ECSF-YOLOv7棉田虫害检测模型。首先,采用EfficientFormerV2作为特征提取网络,以加强网络...棉田虫害的快速检测与准确识别是预防棉田虫害、提高棉花品质的重要前提。针对真实棉田环境下昆虫相似度高、背景干扰严重的问题,该研究提出一种ECSF-YOLOv7棉田虫害检测模型。首先,采用EfficientFormerV2作为特征提取网络,以加强网络的特征提取能力并减少模型参数量;同时,将卷积注意力模块(convolution block attention module,CBAM)嵌入到模型的主干输出端,以增强模型对小目标的特征提取能力并削弱背景干扰;其次,使用GSConv卷积搭建Slim-Neck颈部网络结构,在减少模型参数量的同时保持模型的识别精度;最后,采用Focal-EIOU(focal and efficient IOU loss,Focal-EIOU)作为边界框回归损失函数,加速网络收敛并提高模型的检测准确率。结果表明,改进的ECSF-YOLOv7模型在棉田虫害测试集上的平均精度均值(mean average precision,mAP)为95.71%,检测速度为69.47帧/s。与主流的目标检测模型YOLOv7、SSD、YOLOv5l和YOLOX-m相比,ECSF-YOLOv7模型的mAP分别高出1.43、9.08、1.94、1.52个百分点,并且改进模型具有参数量更小、检测速度更快的优势,可为棉田虫害快速准确检测提供技术支持。展开更多
The source parameters, such as moment tensor, focal mechanism, source time function (STF) and temporal-spatial rupture process, were obtained for the January 26, 2001, India, MS=7.8 earthquake by inverting waveform da...The source parameters, such as moment tensor, focal mechanism, source time function (STF) and temporal-spatial rupture process, were obtained for the January 26, 2001, India, MS=7.8 earthquake by inverting waveform data of 27 GDSN stations with epicentral distances less than 90? Firstly, combining the moment tensor inversion, the spatial distribution of intensity, disaster and aftershocks and the orientation of the fault where the earthquake lies, the strike, dip and rake of the seismogenic fault were determined to be 92? 58?and 62? respectively. That is, this earthquake was a mainly thrust faulting with the strike of near west-east and the dipping direction to south. The seismic moment released was 3.51020 Nm, accordingly, the moment magnitude MW was calculated to be 7.6. And then, 27 P-STFs, 22 S-STFs and the averaged STFs of them were determined respectively using the technique of spectra division in frequency domain and the synthetic seismogram as Greens functions. The analysis of the STFs suggested that the earthquake was a continuous event with the duration time of 19 s, starting rapidly and ending slowly. Finally, the temporal-spatial distribution of the slip on the fault plane was imaged from the obtained P-STFs and S-STFs using an time domain inversion technique. The maximum slip amplitude on the fault plane was about 7 m. The maximum stress drop was 30 MPa, and the average one over the whole rupture area was 7 MPa. The rupture area was about 85 km long in the strike direction and about 60 km wide in the down-dip direction, which, equally, was 51 km deep in the depth direction. The rupture propagated 50 km eastwards and 35 km westwards. The main portion of the rupture area, which has the slip amplitude greater than 0.5 m, was of the shape of an ellipse, its major axis oriented in the slip direction of the fault, which indicated that the rupture propagation direction was in accordance with the fault slip direction. This phenomenon is popular for strike-slip faulting, but rather rare for thrust faulting. The eastern portion of the rupture area above the initiation point was larger than the western portion below the initiation point, which was indicative of the asymmetrical rupture. In other words, the rupturing was kind of unilateral from west to east and from down to up. From the snapshots of the slip-rate variation with time and space, the slip rate reached the largest at the 4th second, that was 0.2 m/s, and the rupture in this period occurred only around the initiation point. At the 6th second, the rupture around the initiation point nearly stopped, and started moving outwards. The velocity of the westward rupture was smaller than that of the eastward rupture. Such rupture behavior like a circle mostly stopped near the 15th second. After the 16th second, only some patches of rupture distributed in the outer region. From the snapshots of the slip variation with time and space, the rupture started at the initiation point and propagated outwards. The main rupture on the area with the slip amplitude greater than 5 m extended unilaterally from west to east and from down to up between the 6th and the 10th seconds, and the western segment extended a bit westwards and downwards between the 11th and the 13th seconds. The whole process lasted about 19 s. The rupture velocity over the whole rupture process was estimated to be 3.3 km/s.展开更多
2016年11月25日在我国新疆克孜勒苏州阿克陶县发生M_S6.7地震(阿克陶M_S6.7地震).我们收集国内外地震资料,对主震及4级以上余震进行了重新定位和震源机制反演,对434次余震进行了双差定位,对主震震源过程进行了反演确定和复杂性分析,并...2016年11月25日在我国新疆克孜勒苏州阿克陶县发生M_S6.7地震(阿克陶M_S6.7地震).我们收集国内外地震资料,对主震及4级以上余震进行了重新定位和震源机制反演,对434次余震进行了双差定位,对主震震源过程进行了反演确定和复杂性分析,并基于反演确定的有限动态源模型估计了此次地震的烈度分布.结果表明:这次地震发生在当地一个近乎东西向展布的小型盆地内,很可能由一条新断层或隐伏断层的活动所致.发震断层近乎直立,近东西向展布,总体上表现为右旋走滑.破裂首先向西扩展,紧接着向东,随后向东西两个方向同时扩展,然后西侧破裂首先停止,东侧破裂继续,最后破裂在东侧停止,整个过程持续一20 s,释放地震矩1.08 X 10^(19)N·m,相当于M_w6.6.破裂过程最终形成两个位错高值区,分别位于初始破裂点的东西两侧,西侧高值区规模较小,东侧区规模较大.根据烈度估计,烈度椭圆长轴方向与主震破裂方向以及余震展布方向一致,最大烈度约为Ⅸ度,主要集中在震中以东很小的区域,Ⅷ度区呈纺锤形,分布于震中东西两侧,Ⅴ至Ⅶ度区呈椭圆形,总体上东侧烈度大于西侧.展开更多
文摘棉田虫害的快速检测与准确识别是预防棉田虫害、提高棉花品质的重要前提。针对真实棉田环境下昆虫相似度高、背景干扰严重的问题,该研究提出一种ECSF-YOLOv7棉田虫害检测模型。首先,采用EfficientFormerV2作为特征提取网络,以加强网络的特征提取能力并减少模型参数量;同时,将卷积注意力模块(convolution block attention module,CBAM)嵌入到模型的主干输出端,以增强模型对小目标的特征提取能力并削弱背景干扰;其次,使用GSConv卷积搭建Slim-Neck颈部网络结构,在减少模型参数量的同时保持模型的识别精度;最后,采用Focal-EIOU(focal and efficient IOU loss,Focal-EIOU)作为边界框回归损失函数,加速网络收敛并提高模型的检测准确率。结果表明,改进的ECSF-YOLOv7模型在棉田虫害测试集上的平均精度均值(mean average precision,mAP)为95.71%,检测速度为69.47帧/s。与主流的目标检测模型YOLOv7、SSD、YOLOv5l和YOLOX-m相比,ECSF-YOLOv7模型的mAP分别高出1.43、9.08、1.94、1.52个百分点,并且改进模型具有参数量更小、检测速度更快的优势,可为棉田虫害快速准确检测提供技术支持。
基金973 Project (G1998040705) from Ministry of Science and Technology, P. R. China the National Science Foundation of China under grant No.49904004.
文摘The source parameters, such as moment tensor, focal mechanism, source time function (STF) and temporal-spatial rupture process, were obtained for the January 26, 2001, India, MS=7.8 earthquake by inverting waveform data of 27 GDSN stations with epicentral distances less than 90? Firstly, combining the moment tensor inversion, the spatial distribution of intensity, disaster and aftershocks and the orientation of the fault where the earthquake lies, the strike, dip and rake of the seismogenic fault were determined to be 92? 58?and 62? respectively. That is, this earthquake was a mainly thrust faulting with the strike of near west-east and the dipping direction to south. The seismic moment released was 3.51020 Nm, accordingly, the moment magnitude MW was calculated to be 7.6. And then, 27 P-STFs, 22 S-STFs and the averaged STFs of them were determined respectively using the technique of spectra division in frequency domain and the synthetic seismogram as Greens functions. The analysis of the STFs suggested that the earthquake was a continuous event with the duration time of 19 s, starting rapidly and ending slowly. Finally, the temporal-spatial distribution of the slip on the fault plane was imaged from the obtained P-STFs and S-STFs using an time domain inversion technique. The maximum slip amplitude on the fault plane was about 7 m. The maximum stress drop was 30 MPa, and the average one over the whole rupture area was 7 MPa. The rupture area was about 85 km long in the strike direction and about 60 km wide in the down-dip direction, which, equally, was 51 km deep in the depth direction. The rupture propagated 50 km eastwards and 35 km westwards. The main portion of the rupture area, which has the slip amplitude greater than 0.5 m, was of the shape of an ellipse, its major axis oriented in the slip direction of the fault, which indicated that the rupture propagation direction was in accordance with the fault slip direction. This phenomenon is popular for strike-slip faulting, but rather rare for thrust faulting. The eastern portion of the rupture area above the initiation point was larger than the western portion below the initiation point, which was indicative of the asymmetrical rupture. In other words, the rupturing was kind of unilateral from west to east and from down to up. From the snapshots of the slip-rate variation with time and space, the slip rate reached the largest at the 4th second, that was 0.2 m/s, and the rupture in this period occurred only around the initiation point. At the 6th second, the rupture around the initiation point nearly stopped, and started moving outwards. The velocity of the westward rupture was smaller than that of the eastward rupture. Such rupture behavior like a circle mostly stopped near the 15th second. After the 16th second, only some patches of rupture distributed in the outer region. From the snapshots of the slip variation with time and space, the rupture started at the initiation point and propagated outwards. The main rupture on the area with the slip amplitude greater than 5 m extended unilaterally from west to east and from down to up between the 6th and the 10th seconds, and the western segment extended a bit westwards and downwards between the 11th and the 13th seconds. The whole process lasted about 19 s. The rupture velocity over the whole rupture process was estimated to be 3.3 km/s.
文摘2016年11月25日在我国新疆克孜勒苏州阿克陶县发生M_S6.7地震(阿克陶M_S6.7地震).我们收集国内外地震资料,对主震及4级以上余震进行了重新定位和震源机制反演,对434次余震进行了双差定位,对主震震源过程进行了反演确定和复杂性分析,并基于反演确定的有限动态源模型估计了此次地震的烈度分布.结果表明:这次地震发生在当地一个近乎东西向展布的小型盆地内,很可能由一条新断层或隐伏断层的活动所致.发震断层近乎直立,近东西向展布,总体上表现为右旋走滑.破裂首先向西扩展,紧接着向东,随后向东西两个方向同时扩展,然后西侧破裂首先停止,东侧破裂继续,最后破裂在东侧停止,整个过程持续一20 s,释放地震矩1.08 X 10^(19)N·m,相当于M_w6.6.破裂过程最终形成两个位错高值区,分别位于初始破裂点的东西两侧,西侧高值区规模较小,东侧区规模较大.根据烈度估计,烈度椭圆长轴方向与主震破裂方向以及余震展布方向一致,最大烈度约为Ⅸ度,主要集中在震中以东很小的区域,Ⅷ度区呈纺锤形,分布于震中东西两侧,Ⅴ至Ⅶ度区呈椭圆形,总体上东侧烈度大于西侧.
