加窗时空分析法实现三维测量

Fast 3D Measurement through Windowed Space-Time Analysis

分析了传统时空分析法空间和时间成本,发现过多无效数据的处理是导致计算效率低下的原因,提出了加窗时空分析法.该方法通过对时间序列的数据进行加窗截取,缩小了数据计算范围,从而减轻了内存负担,提高了计算效率.同时,小窗口的浮动特性使得延迟计算时间缩小到1/2窗口值,并且可实现随着图像的获取进行逐帧计算.对3个实验物体进行三维测量,曲面实验结果精细正确.在与传统时空分析法的定量比较中,该方法结果与传统方法结果在距离阈值10-6下的正确率和回调率高达99.8%,且计算时间仅为传统方法的2%.因此,加窗时空分析法使用较小的计算成本实现了高质量的计算结果.

Space-time analysis is one of the most robust methods to measure 3D shape. However, it has some disadvantages, such as large storage burden, high computational load, long waiting time. A fast space-time analysis was proposed which applies the windowed data instead of the whole time series data. The new method greatly reduces the computational load, storage burden and waiting time. The experi- ments on three test objects were implemented. The surfaces using the proposed method are very fine, even the small texture and wave on the surface are also reconstructed. In the quantitative comparison with the traditional space-time analysis, the accuracies and recalls are as high as 99.8%. what is move, the efficien- cy is enhanced rapidly. The computational time is just 2% of the traditional one. The experiment results confirm that the windowed space-time analysis can obtain high quality computing result while using very low computing cost.