Please help me correct the grammar for this paper:
Experiments with UAS Imagery
Introduction
The article describes the advantages and disadvantages of Unmanned Airborne Systems (UAS) comparing to airborne mapping systems. The main purpose is focusing on the geo-referencing of the USA-acquired imagery. The latest technology provides two similar, reliable results and confirming indirect and direct geo-referencing to reduce the field work, Ground Control Points (GCPs) surveying, but appropriate with other sensors, such as active imaging technology, LiDAR, further extending UAS application potential. Also, UAS image geolocation are approached by three different positioning: based on using dual-frequency GPS data, uses the single frequency code only and employs indirect image geo-referencing based on aerial triangulation (AT) using ground controls.
Approach
UAS is currently of high interest in the mapping community. Data collection was performed using a Bergen octocopter, flown in fully autonomous mode based on the planned coordinates of waypoints. To support for UAS platform positioning and the images were acquired at the shortest time, the following sensors were installed on the octocopter: Nikon D800 36 Mpix camera, Dual-frequency GPS receiver and Single frequency code GPS receiver. Most of the results obtained by indirect georeferencing require more work but more comprehensive and accurate than direct georeferencing in survey design. Indirect georeferencing uses ground control targets placed in the survey area with known coordinates from GPS. This method can observe the overlapping images, solve the exterior orientation and make an aligned and georeferenced set of images.
High-performance aerial mapping systems use GPS/IMU-based direct georeferencing. The main advantage of direct measurement of image position and orientation during the flight is the lack of field surveys to collect GCPs. Besides the requirements for the sensors performance, other issues apply to direct geo-referencing, such as time synchronization between all the sensors as well as boresight misalignment.
How to measure and compare the accuracy results? A small computer, fit-PC2, was mounted on the UAS for the parameter setup and GPS data recording. Also, there are two aspects: first, lever arm offsets can be either measured or estimated and considered during adjustment, but not necessary for indirect geo-referencing because GPS data is not used. In contrast, the accuracy of post-processed, the second important aspect of sensor integration is the time synchronization which allows linking images with appropriate GPS recorded positions. The GPS time for every camera shutter release is known at around 1 ms accuracy.
Results and discussion
Based on three different approaches, the results shows that the position using dual-frequency GPS receiver and the bundle clock adjustment to be similar. The major results of the bundle adjustment give the list of RMSE of the air controls and ground check points. The result said that total accuracy of 0.3 m for image positions seem to be much worse that compare to AT based on GCPs (4.7 cm). The lack of IMU data and not including horizontal offsets let us know that the larger horizontal than vertical RMSE for air controls. Horizontal shifts need to be considered for the camera position due to platform rotation related to non-fixed-wing UAS. Large numbers of the GCPs and tie-points as along with small errors suggest that the geolocation of images was correctly estimated. For horizontal position, the State Plane Coordinate System (SPCS) was used and vertical coordinates were converted into the geodetic heights using GEOID12A model. This gave the common base for direct comparison of image geolocation.
Testing three image positions based on height profile and horizontal trajectory, same time range, 5 Hz dual frequency GPS date, note that height of reduced already by vertical offset 35cm between antenna and camera center. If we look closer each single position from the bundle adjustment seems to be slightly off and non-smooth segment was not observed during flight. In horizontal position, the largest differences about 1m. The different results are in level arm offset. Comparing vertical position, there is a shift of the mean value equal about 0.3 m and the error might be come from camera self-calibration and wrong estimation of the principle distance. Since there is no air control, camera error may affect the image positions adjusted in aerial triangulation (AT). Assum that the results obtained from dual frequency GPS data air AT are the reference value. The horizontal error is equal to almost 7m and much more than device specified 3 m.
Conclusion
Model technique methods can automate feature matching between overlapping imagery. This lab helps me understand a study performed to investigate geo-referencing accuracy of UAS imagery using different methods of either low or high grade GPS and indirect geo-referencing with control target. The author provides a better method to increase accurate position especial in horizontal direction by using high grade GPS receivers. Improvement of MEMS based IMU sensors as well as observed LiDAR sensors will further advance the use of UAS mapping technology. In the future, UAS mapping products should use high-end airborne system to improve quality. Beside that, smaller UASs equipment will cost more compare to old low cost devices. So hardware price is still a problem need to be mention