通过研究,提出了一种基于不确定性建模的中文场景文本编辑(Chinese scene text editing,CSTE)方法,并发现了1种有效的技术解决方案.该方法通过不确定性引导的调整机制优化预测噪声,提升噪声估计准确性,从而增强生成文本的清晰度和结构...通过研究,提出了一种基于不确定性建模的中文场景文本编辑(Chinese scene text editing,CSTE)方法,并发现了1种有效的技术解决方案.该方法通过不确定性引导的调整机制优化预测噪声,提升噪声估计准确性,从而增强生成文本的清晰度和结构完整性.同时,通过过滤文本和图像特征中的无关信息,提高了跨模态对齐能力,实现了文本与背景纹理的融合.展开更多
The quality of the micro-mechanical machining outcome depends significantly on the tracking performance of the miniaturized linear motor drive precision stage. The tracking behavior of a direct drive design is prone t...The quality of the micro-mechanical machining outcome depends significantly on the tracking performance of the miniaturized linear motor drive precision stage. The tracking behavior of a direct drive design is prone to uncertainties such as model parameter variations and disturbances. Robust optimal tracking controller design for this kind of precision stages with mass and damping ratio uncertainties was researched. The mass and damping ratio uncertainties were modeled as the structured parametric uncertainty model. An identification method for obtaining the parametric uncertainties was developed by using unbiased least square technique. The instantaneous frequency bandwidth of the external disturbance signals was analyzed by using short time Fourier transform technique. A two loop tracking control strategy that combines the p-synthesis and the disturbance observer （DOB） techniques was proposed. The p-synthesis technique was used to design robust optimal controllers based on structured uncertainty models. By complementing the/z controller, the DOB was applied to further improving the disturbance rejection performance. To evaluate the positioning performance of the proposed control strategy, the comparative experiments were conducted on a prototype micro milling machine among four control schemes： the proposed two-loop tracking control, the single loop μ control, the PID control and the PID with DOB control. The disturbance rejection performances, the root mean square （RMS） tracking errors and the performance robustness of different control schemes were studied. The results reveal that the proposed control scheme has the best positioning performance. It reduces the maximal errors caused by disturbance forces such as friction force by 60% and the RMS errors by 63.4% compared with the PID control. Compared to PID with DOB control, it reduces the RMS errors by 29.6%.展开更多
文摘通过研究,提出了一种基于不确定性建模的中文场景文本编辑(Chinese scene text editing,CSTE)方法,并发现了1种有效的技术解决方案.该方法通过不确定性引导的调整机制优化预测噪声,提升噪声估计准确性,从而增强生成文本的清晰度和结构完整性.同时,通过过滤文本和图像特征中的无关信息,提高了跨模态对齐能力,实现了文本与背景纹理的融合.
基金Project(50875257) supported by the National Natural Science Foundation of China
文摘The quality of the micro-mechanical machining outcome depends significantly on the tracking performance of the miniaturized linear motor drive precision stage. The tracking behavior of a direct drive design is prone to uncertainties such as model parameter variations and disturbances. Robust optimal tracking controller design for this kind of precision stages with mass and damping ratio uncertainties was researched. The mass and damping ratio uncertainties were modeled as the structured parametric uncertainty model. An identification method for obtaining the parametric uncertainties was developed by using unbiased least square technique. The instantaneous frequency bandwidth of the external disturbance signals was analyzed by using short time Fourier transform technique. A two loop tracking control strategy that combines the p-synthesis and the disturbance observer （DOB） techniques was proposed. The p-synthesis technique was used to design robust optimal controllers based on structured uncertainty models. By complementing the/z controller, the DOB was applied to further improving the disturbance rejection performance. To evaluate the positioning performance of the proposed control strategy, the comparative experiments were conducted on a prototype micro milling machine among four control schemes： the proposed two-loop tracking control, the single loop μ control, the PID control and the PID with DOB control. The disturbance rejection performances, the root mean square （RMS） tracking errors and the performance robustness of different control schemes were studied. The results reveal that the proposed control scheme has the best positioning performance. It reduces the maximal errors caused by disturbance forces such as friction force by 60% and the RMS errors by 63.4% compared with the PID control. Compared to PID with DOB control, it reduces the RMS errors by 29.6%.
