Accuracy analysis of a multi-camera imaging system mounted on a UAS after image mosaicking

In recent years, Unmanned Aerial System (UAS) has been applied to collect remote sensing data for category mapping, disaster investigation, vegetation monitoring and etc. However, due to the low payload and short operation time that reduces the mobility and image collection efficiency, only small frame camera is mounted for small area mapping purpose. In this study, one nadir and four oblique view cameras composed multi-camera imaging system is mounted on a high-payload UAS, which has capability to increase the field of view (FOV) for large ground coverage area topographic mapping. To reduce the number of unknown parameters and images, the original multi images taken in the same time are mosaicked into one large frame image through a modified projective transformation (MPT) model with two systematic error correction schemes. The coefficients of MPT are estimated from an indoor camera calibration filed, which calibrates the interior orientation parameters (IOPs) of each camera and relative orientation parameters (ROPs) among four oblique cameras and nadir camera. The systematic errors are considered since the environment of camera calibration field and flight conditions are different, the scale and displacement errors during image mosaicking are thus analyzed and corrected. The experimental result shows that the internal accuracy (average length of residuals) of the mosaicked images can achieve sub-pixel accuracy. The comparison of topographic mapping accuracy among mosaicked image and original images were conducted through rigorous aerial triangulation. The results show the mosaicked images could achieve the mapping of 1/1000 topographic standard.