为使水果采摘机器人在复杂情况下如树叶遮挡、果实目标尺度变化大等情况能准确地检测出水果,提出一种YOLO(you only look once)改进模型与NMS(non-maximum suppression)改进算法的目标检测方法。首先,对传统YOLO深度卷积神经网络架构进...为使水果采摘机器人在复杂情况下如树叶遮挡、果实目标尺度变化大等情况能准确地检测出水果,提出一种YOLO(you only look once)改进模型与NMS(non-maximum suppression)改进算法的目标检测方法。首先,对传统YOLO深度卷积神经网络架构进行改进,设计一种更细化的SPP5(spatial pyramid pooling)特征融合网络模块,强化特征图多重感受野信息的融合,并基于此模块提出一种YOLOv4-SPP2-5模型,在标准YOLOv4网络中跨层添加并改进SPP层,重新分布池化核大小,增强感受野范围,从而降低目标误检率;其次,提出一种Greedy-Confluence的NMS改进算法,通过对高度接近的检测框直接抑制和对重叠检测框综合考虑距离交并比DIOU(distance-intersection over union)和加权接近度WP(weighted proximity)的方法,均衡NMS的计算消耗并减少检测框的错误抑制,从而提高遮挡、重叠物体的检测精度;最后,分别对改进方法进行性能测试,验证方法的可行性,随后制作水果检测数据集并进行格式转换和标签标注,然后采用数据增强技术对训练数据进行扩充,并使用K-means++聚类方法获取先验锚定框,在计算机上进行了水果检测实验。结果表明,基于改进YOLO网络及改进NMS的水果检测方法在准确率方面有显著的提高,平均精度均值(mean average precision,MAP)在YOLOv4上达到了96.65%,较原网络提升1.70%,并且实时性也得到了保证,在测试设备上达到了39.26帧/s。展开更多
Various Ionic and molecular activities in the extraceUular environment are vital to plant cell physiological processes. A noninvasive microsensing system （NMS） based on either the scanning ion-selective electrode te...Various Ionic and molecular activities in the extraceUular environment are vital to plant cell physiological processes. A noninvasive microsensing system （NMS） based on either the scanning ion-selective electrode technique （SIET） or the scanning polarographlc electrode technique （SPET） is able to obtain information regarding the transportation of various Ions/molecules in Intact samples under normal physiological conditions. The two-probe simultaneous test system （2STS） Is an Integrated system composed of SIET, SPET, and a Xu-Kunkel sampling protocol. In the present study, 2STS was able to simultaneously measure fluxes of H^＋ and O2 of the Uly （Lillum Iongiflorum Thunb. cv. Ace） pollen tube while avoiding interference between the two probes. The results Indicate that the proton fluxes were effluxes, whereas the oxygen fluxes were Influxes, and they were closely correlated to each other surrounding the constitutive alkaline band region. Specifically, when the proton effluxes increased, the oxygen Influxes also increased. Therefore, the hypothesis of condensed active mitochondria existing in the alkalized area of the pollen tube proposed by Hepler＇s group is supported.展开更多
Twenty-five tree species indigenous to Guangdong Province were chosen in this study to portray their distribution patterns in relation to environmental factors. Both data of species distribution and environmental fact...Twenty-five tree species indigenous to Guangdong Province were chosen in this study to portray their distribution patterns in relation to environmental factors. Both data of species distribution and environmental factors were tabulated based on a digitized map of Guangdong Province gridded at 0.5° latitude × 0.5° longitude. Grid-based diversity was mapped using DMAP, a distribution mapping program, and horizontal patterns were assessed using Kruskal-Wallis tests. The diversity center of the indige- nous tree species under study is located north of 23° N. These tree species exhibit significant latitudinal variation (P = 0.007 4), but no significant longitudinal difference (P = 0.052 2). Non-metric Multidimensional Scaling (NMS) identified five different ecological species groups, while Canonical Correspondence Analysis (CCA) showed the distribution of tree species along each of the five envi- ronmental gradients. An understanding of the environmental correlates of distribution patterns has great implication for the introduc- tion of the indigenous tree species for afforestation.展开更多
文摘为使水果采摘机器人在复杂情况下如树叶遮挡、果实目标尺度变化大等情况能准确地检测出水果,提出一种YOLO(you only look once)改进模型与NMS(non-maximum suppression)改进算法的目标检测方法。首先,对传统YOLO深度卷积神经网络架构进行改进,设计一种更细化的SPP5(spatial pyramid pooling)特征融合网络模块,强化特征图多重感受野信息的融合,并基于此模块提出一种YOLOv4-SPP2-5模型,在标准YOLOv4网络中跨层添加并改进SPP层,重新分布池化核大小,增强感受野范围,从而降低目标误检率;其次,提出一种Greedy-Confluence的NMS改进算法,通过对高度接近的检测框直接抑制和对重叠检测框综合考虑距离交并比DIOU(distance-intersection over union)和加权接近度WP(weighted proximity)的方法,均衡NMS的计算消耗并减少检测框的错误抑制,从而提高遮挡、重叠物体的检测精度;最后,分别对改进方法进行性能测试,验证方法的可行性,随后制作水果检测数据集并进行格式转换和标签标注,然后采用数据增强技术对训练数据进行扩充,并使用K-means++聚类方法获取先验锚定框,在计算机上进行了水果检测实验。结果表明,基于改进YOLO网络及改进NMS的水果检测方法在准确率方面有显著的提高,平均精度均值(mean average precision,MAP)在YOLOv4上达到了96.65%,较原网络提升1.70%,并且实时性也得到了保证,在测试设备上达到了39.26帧/s。
基金Supported by the National Natural Science Foundation of China （30170552）, Beijing Natural Science Foundation （5042004）, Beijing Education Committee Science and Technology Development Project （KM200410028015） and Applicable Electronics Inc. （Forestdale, MA, USA）.
文摘Various Ionic and molecular activities in the extraceUular environment are vital to plant cell physiological processes. A noninvasive microsensing system （NMS） based on either the scanning ion-selective electrode technique （SIET） or the scanning polarographlc electrode technique （SPET） is able to obtain information regarding the transportation of various Ions/molecules in Intact samples under normal physiological conditions. The two-probe simultaneous test system （2STS） Is an Integrated system composed of SIET, SPET, and a Xu-Kunkel sampling protocol. In the present study, 2STS was able to simultaneously measure fluxes of H^＋ and O2 of the Uly （Lillum Iongiflorum Thunb. cv. Ace） pollen tube while avoiding interference between the two probes. The results Indicate that the proton fluxes were effluxes, whereas the oxygen fluxes were Influxes, and they were closely correlated to each other surrounding the constitutive alkaline band region. Specifically, when the proton effluxes increased, the oxygen Influxes also increased. Therefore, the hypothesis of condensed active mitochondria existing in the alkalized area of the pollen tube proposed by Hepler＇s group is supported.
基金Key Research Program of Guangdong Province (Grant No. 2002C20703) and Key Research Program of Guangdong Provincial Department ofForestry (Grant No. 2002-12)
文摘Twenty-five tree species indigenous to Guangdong Province were chosen in this study to portray their distribution patterns in relation to environmental factors. Both data of species distribution and environmental factors were tabulated based on a digitized map of Guangdong Province gridded at 0.5° latitude × 0.5° longitude. Grid-based diversity was mapped using DMAP, a distribution mapping program, and horizontal patterns were assessed using Kruskal-Wallis tests. The diversity center of the indige- nous tree species under study is located north of 23° N. These tree species exhibit significant latitudinal variation (P = 0.007 4), but no significant longitudinal difference (P = 0.052 2). Non-metric Multidimensional Scaling (NMS) identified five different ecological species groups, while Canonical Correspondence Analysis (CCA) showed the distribution of tree species along each of the five envi- ronmental gradients. An understanding of the environmental correlates of distribution patterns has great implication for the introduc- tion of the indigenous tree species for afforestation.
