绵羊的姿态与其健康及福利密切相关。随着智能化畜牧业需求的增长,自动、准确地检测绵羊姿态尤为尤为重要。本研究提出基于Mask R-CNN基准网络的新型RDS-Mask R-CNN绵羊姿态检测算法,以Res2Net101作为特征提取网络,同时引入可变形卷积(D...绵羊的姿态与其健康及福利密切相关。随着智能化畜牧业需求的增长,自动、准确地检测绵羊姿态尤为尤为重要。本研究提出基于Mask R-CNN基准网络的新型RDS-Mask R-CNN绵羊姿态检测算法,以Res2Net101作为特征提取网络,同时引入可变形卷积(Deformable convolution network,DCN),以更精准捕捉绵羊在不同位置的姿态特征,并运用软非极大值抑制(Soft non-maximum suppression,Soft NMS)算法实现重叠实例目标的准确分割。结果表明:1)目标检测框架算法对比:与该领域最经典的YOLOv3和Faster R-CNN相比,改进的算法在平均精度均值(Mean average precision,mAP)上分别提升了16.68%和8.64%;2)不同改进策略的算法对比:改进算法相较于基准网络,边界框平均精度均值(Bounding box mean average precision,Bbox mAP)提高6.21%,分割平均精度均值(Segmentation mean average precision,Segm mAP)提高6.61%,分别达到87.34%和81.50%;3)相较于Mask R-CNN,改进模型在识别绵羊站立与躺卧姿态时边界框平均精度(Bounding box average precision,Bbox AP)分别提高了6.84%和5.58%,分割平均精度(Segmentation average precision,Segm AP)分别提高了7.25%和5.17%;4)模型可解释性可视化结果表明RDS-Mask R-CNN能精准捕获绵羊站立和躺卧姿态关键部位深度特征,表明模型自动检测可行且具有可解释性。综上,本研究提出的RDS-Mask R-CNN算法,有效提升了绵羊姿态检测的精准度,为智慧养殖提供了技术支撑。展开更多
Data from rain Drop Size Distributions gathered on five sites in Africa as well as those of the pilot site in Kourou (French Guyana, South America), located in different climatic zones, and collected by two types of d...Data from rain Drop Size Distributions gathered on five sites in Africa as well as those of the pilot site in Kourou (French Guyana, South America), located in different climatic zones, and collected by two types of disdrometer (the impact JW RD-69 disdrometer and the Optical Spectro-Pluviometer, OSP) are used to study the consistency of the reflectivity factor-rain rate at the ground (Z-R) relationship variability. The results clearly confirm that the relationship Z-R knows a large spatial variability, from a type of precipitation to another and within the same precipitation regardless the type of disdrometer used for DSD measurements. Base on the similarity of the relations reflectivity factor-rain rate and ratio median volume diameter over the total number of drops-rain rate, the variability of the Z-R coefficients (A, b) through the simultaneously implication of the size and number of drops which characterize the DSD was exhibited. It was shown that the relationships A-α and b-β designed to understand the involvement of parameters D0 and NT of DSD in the variability of the relationship Z-R are similar regardless the types of disdrometer used. However, the relations A-α in the Sahelian region appear to deviate from those of Guinean, equatorial and Soudanian zones. The plausible reasons were discussed.展开更多
文摘绵羊的姿态与其健康及福利密切相关。随着智能化畜牧业需求的增长,自动、准确地检测绵羊姿态尤为尤为重要。本研究提出基于Mask R-CNN基准网络的新型RDS-Mask R-CNN绵羊姿态检测算法,以Res2Net101作为特征提取网络,同时引入可变形卷积(Deformable convolution network,DCN),以更精准捕捉绵羊在不同位置的姿态特征,并运用软非极大值抑制(Soft non-maximum suppression,Soft NMS)算法实现重叠实例目标的准确分割。结果表明:1)目标检测框架算法对比:与该领域最经典的YOLOv3和Faster R-CNN相比,改进的算法在平均精度均值(Mean average precision,mAP)上分别提升了16.68%和8.64%;2)不同改进策略的算法对比:改进算法相较于基准网络,边界框平均精度均值(Bounding box mean average precision,Bbox mAP)提高6.21%,分割平均精度均值(Segmentation mean average precision,Segm mAP)提高6.61%,分别达到87.34%和81.50%;3)相较于Mask R-CNN,改进模型在识别绵羊站立与躺卧姿态时边界框平均精度(Bounding box average precision,Bbox AP)分别提高了6.84%和5.58%,分割平均精度(Segmentation average precision,Segm AP)分别提高了7.25%和5.17%;4)模型可解释性可视化结果表明RDS-Mask R-CNN能精准捕获绵羊站立和躺卧姿态关键部位深度特征,表明模型自动检测可行且具有可解释性。综上,本研究提出的RDS-Mask R-CNN算法,有效提升了绵羊姿态检测的精准度,为智慧养殖提供了技术支撑。
文摘Data from rain Drop Size Distributions gathered on five sites in Africa as well as those of the pilot site in Kourou (French Guyana, South America), located in different climatic zones, and collected by two types of disdrometer (the impact JW RD-69 disdrometer and the Optical Spectro-Pluviometer, OSP) are used to study the consistency of the reflectivity factor-rain rate at the ground (Z-R) relationship variability. The results clearly confirm that the relationship Z-R knows a large spatial variability, from a type of precipitation to another and within the same precipitation regardless the type of disdrometer used for DSD measurements. Base on the similarity of the relations reflectivity factor-rain rate and ratio median volume diameter over the total number of drops-rain rate, the variability of the Z-R coefficients (A, b) through the simultaneously implication of the size and number of drops which characterize the DSD was exhibited. It was shown that the relationships A-α and b-β designed to understand the involvement of parameters D0 and NT of DSD in the variability of the relationship Z-R are similar regardless the types of disdrometer used. However, the relations A-α in the Sahelian region appear to deviate from those of Guinean, equatorial and Soudanian zones. The plausible reasons were discussed.
