*3.3. Close-Range Photogrammetry*

The geometry of the site is rather peculiar, with a huge reservoir in front of the hillside (Figure 1), with many crossing hydroelectric power lines from one side to the other, which hamper the exploitation of close-range imagery with RPAS–UAVs (remotely piloted aircraft system–unmanned aerial vehicles). The approximate dimension of the interest hillside displayed in Figure 1 (yellow box) is 700 m wide and 200 m high. In addition, pictures were taken not from ideally planned positions but realistically ideal positions on-site, after defining some minimum constraints such as ideal ground sampling distance (GSD) e.g., following roads parallel to the hillside, and from surrounding hills.

For the CRP campaign, two 21 Mpix full-frame cameras were used mounted on stable tripods: Canon EOS D5 Mark II with a normal Canon EF 50 mm f/1.4 USM lens, and a Canon 1Ds Mark III with a telephoto Canon EF 200 mm f/2,8 L II USMlens (Figure 4). The first camera was used to ge<sup>t</sup> the whole area, summing 280 normal and convergen<sup>t</sup> imagery; the second camera was used to measure only three areas with the fixed on-the-rock 15 check points, summing 404 imagery. The data acquisition was planned to achieve a GSD of 3.5 cm. The mean camera-object distance was 512 m. A total of 6 h were devoted to taking images on site.

**Figure 4.** Images acquired from the same spatial position with two fixed lenses: (**a**) normal 50 mm; (**b**) telephoto 200 mm.

The three-step workflow followed to achieve accuracy better than 3 cm is presented next. First step, each camera was calibrated on the job selecting a subsampled ideal network with convergen<sup>t</sup> and rotated imagery, summing up to 57 images for the 50 mm lens and 47 images for the 200 mm lens. For the normal lens, 8 interior orientation parameters (f, cx, cy, K1, K2, K3, P1, P2) were computed to achieve 0.34 pixel image reprojection error; for the telephoto lens, 4 additional parameters (f, cx, cy, K1) were computed to achieve a 0.50 pixel image reprojection error. Second step, a free bundle block adjustment with 230 imagery (142 normal + 88 tele) was undertaken. This step included the manual measurement of the white spherical targets (Figures 2b and 3) on the imagery, including 4 (visible) control points on pillars of the reference frame and 15 check points (Figure 5). A mean reprojection error of 0.50 pixels was achieved in the photogrammetric adjustment. Third step, absolute orientation by means of a 3D similarity transformation yielding an RMSE (root mean square error) of 1.55 cm.

**Figure 5.** Set of imagery used for the bundle adjustment of the normal and tele cameras and visualization of the measured points on geodetic pillars and check points.
