In multi-LFM signal condition, Radon-Ambiguity Transform (RAT) of the strong LFM component has strong suppression effect on that of the weak LFM component. A method named as Recursive Filtering RAT (RFRAT) algorithm i...In multi-LFM signal condition, Radon-Ambiguity Transform (RAT) of the strong LFM component has strong suppression effect on that of the weak LFM component. A method named as Recursive Filtering RAT (RFRAT) algorithm is proposed for solving this problem. By fully using of the Maximum Likelihood (ML) estimation value of the frequency modulation rate got by RAT, RFRAT can detect the noisy multi-LFM signals out step by step. The merit of this new method is validated by an illustrative example in low Signal-to-Noise-Ratio (SNR) condition.展开更多
The Radon-ambiguity transform （RAT）, although efficient for detecting the linear frequency modulated signals （LFMs）, is troubled by the energy accumulation of noise in low signal-to-noise ratio （SNR）. A secondor...The Radon-ambiguity transform （RAT）, although efficient for detecting the linear frequency modulated signals （LFMs）, is troubled by the energy accumulation of noise in low signal-to-noise ratio （SNR）. A secondorder difference （SOD） method is proposed to treat with this problem. In the SOD method, the optimal search step and difference step are derived from the LFM rate resolution formula. The sharpness of the peaks of RAT is measured by curvature, and the sharpness, but not the magnitude of the peaks, is used to detect the LFMs. The SOD method removes the noise energy accumulation and reserves the drastically changing components integrally; thus, it improves the detection probability of LFMs in low SNR. The expected performance of the new method is verified by 100 Monte Carlo simulations.展开更多
针对地图综合中建筑多边形化简方法依赖人工规则、自动化程度低且难以利用已有化简成果的问题,本文提出了一种基于Transformer机制的建筑多边形化简模型。该模型首先把建筑多边形映射至一定范围的网格空间,将建筑多边形的坐标串表达为...针对地图综合中建筑多边形化简方法依赖人工规则、自动化程度低且难以利用已有化简成果的问题,本文提出了一种基于Transformer机制的建筑多边形化简模型。该模型首先把建筑多边形映射至一定范围的网格空间,将建筑多边形的坐标串表达为网格序列,从而获取建筑多边形化简前后的Token序列,构建出建筑多边形化简样本对数据;随后采用Transformer架构建立模型,基于样本数据利用模型的掩码自注意力机制学习点序列之间的依赖关系,最终逐点生成新的简化多边形,从而实现建筑多边形的化简。在训练过程中,模型使用结构化的样本数据,设计了忽略特定索引的交叉熵损失函数以提升化简质量。试验设计包括主试验与泛化验证两部分。主试验基于洛杉矶1∶2000建筑数据集,分别采用0.2、0.3和0.5 mm 3种网格尺寸对多边形进行编码,实现了目标比例尺为1∶5000与1∶10000的化简。试验结果表明,在0.3 mm的网格尺寸下模型性能最优,验证集上的化简结果与人工标注的一致率超过92.0%,且针对北京部分区域的建筑多边形数据的泛化试验验证了模型的迁移能力;与LSTM模型的对比分析显示,在参数规模相近的条件下,LSTM模型无法形成有效收敛,并生成可用结果。本文证实了Transformer在处理空间几何序列任务中的潜力,且能够有效复用已有化简样本,为智能建筑多边形化简提供了具有工程实用价值的途径。展开更多
基金Supported by the National 973 Program(No.973-1-12)
文摘In multi-LFM signal condition, Radon-Ambiguity Transform (RAT) of the strong LFM component has strong suppression effect on that of the weak LFM component. A method named as Recursive Filtering RAT (RFRAT) algorithm is proposed for solving this problem. By fully using of the Maximum Likelihood (ML) estimation value of the frequency modulation rate got by RAT, RFRAT can detect the noisy multi-LFM signals out step by step. The merit of this new method is validated by an illustrative example in low Signal-to-Noise-Ratio (SNR) condition.
基金supported by the Program for New Century Excellent Talents in University, Ministry of Education (NCET-05-0803)
文摘The Radon-ambiguity transform （RAT）, although efficient for detecting the linear frequency modulated signals （LFMs）, is troubled by the energy accumulation of noise in low signal-to-noise ratio （SNR）. A secondorder difference （SOD） method is proposed to treat with this problem. In the SOD method, the optimal search step and difference step are derived from the LFM rate resolution formula. The sharpness of the peaks of RAT is measured by curvature, and the sharpness, but not the magnitude of the peaks, is used to detect the LFMs. The SOD method removes the noise energy accumulation and reserves the drastically changing components integrally; thus, it improves the detection probability of LFMs in low SNR. The expected performance of the new method is verified by 100 Monte Carlo simulations.
文摘针对地图综合中建筑多边形化简方法依赖人工规则、自动化程度低且难以利用已有化简成果的问题,本文提出了一种基于Transformer机制的建筑多边形化简模型。该模型首先把建筑多边形映射至一定范围的网格空间,将建筑多边形的坐标串表达为网格序列,从而获取建筑多边形化简前后的Token序列,构建出建筑多边形化简样本对数据;随后采用Transformer架构建立模型,基于样本数据利用模型的掩码自注意力机制学习点序列之间的依赖关系,最终逐点生成新的简化多边形,从而实现建筑多边形的化简。在训练过程中,模型使用结构化的样本数据,设计了忽略特定索引的交叉熵损失函数以提升化简质量。试验设计包括主试验与泛化验证两部分。主试验基于洛杉矶1∶2000建筑数据集,分别采用0.2、0.3和0.5 mm 3种网格尺寸对多边形进行编码,实现了目标比例尺为1∶5000与1∶10000的化简。试验结果表明,在0.3 mm的网格尺寸下模型性能最优,验证集上的化简结果与人工标注的一致率超过92.0%,且针对北京部分区域的建筑多边形数据的泛化试验验证了模型的迁移能力;与LSTM模型的对比分析显示,在参数规模相近的条件下,LSTM模型无法形成有效收敛,并生成可用结果。本文证实了Transformer在处理空间几何序列任务中的潜力,且能够有效复用已有化简样本,为智能建筑多边形化简提供了具有工程实用价值的途径。
