研究目的
To verify that when the temperature of image sensor varies during the acquisition, image deformation induced by the temperature change is quantifiable, modelisable and correctable, and to improve the geometric accuracy of photogrammetric reconstruction.
研究成果
The thermal deformation is reproducible and can be modeled and corrected, leading to improved photogrammetric accuracy. Focal length variation is consistent (0.4-0.5 μm/°C), and accuracy improvements of up to 1.5 times in terrestrial and 1.4 times in aerial configurations are achieved. Further work is needed on long-term stability and principal point variations.
研究不足
The study does not fully address the influence of external temperature; variations in translation and rotation parameters over time are not fully explained; long-term stability of thermal effects requires further investigation; the number of datasets for bundle adjustment is insufficient for comprehensive analysis.
1：Experimental Design and Method Selection:
The study uses two main methods: 2D image correlation and bundle adjustment to model thermal deformations. A 4-parameter 2D spatial similarity transformation is employed for simplicity, with polynomial models for finer corrections.
2：Sample Selection and Data Sources:
Experiments are conducted using the IGN lightweight metric camera (CamLight) with a CMOSIS CMV20000 sensor. Datasets include indoor and outdoor acquisitions with controlled temperature variations.
3：List of Experimental Equipment and Materials:
Equipment includes the CamLight camera, temperature sensor, Peltier cooler for cooling, tripod, textured wall for correlation, calibration field with GCPs surveyed by total station, and UAV for aerial data.
4：Experimental Procedures and Operational Workflow:
Type I acquisitions involve static camera imaging with increasing temperature; Type II uses cooling to minimize temperature change. Image sequences are captured, and deformation maps are generated by comparing images to a reference. Bundle adjustment is used for parameter estimation.
5：Data Analysis Methods:
Least squares solver for parameter estimation, image correlation for displacement maps, and statistical analysis of reprojection errors and residuals.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
