PSI - Issue 17

Behzad V. Farahani et al. / Procedia Structural Integrity 17 (2019) 712–717 Behzad V. Farahani et al./ Structural Integrity Procedia 00 (2019) 000 – 000

715

4

Table 1: Experimental test configurations. System velocity, Camera 32.00 mm/s 8.00 fps

framerate, Total length

No. images

System Resolution

2561.00 mm

728.00

3.52 mm/frame

Fig. 4-a presents an image captured by the camera. The captured images must be transformed into a binary image through image processing. This task has been accomplished converting the RGB images into CIELAB colour space, in which the lightness value, ∗ , represents the darkest black at ∗ = 0 , and the brightest white at ∗ = 100 . The colour channels, ∗ and ∗ , signify true neutral gray values at ∗ = 0 and ∗ = 0 . Smaller values of ∗ denote the green colour. The points were filtered according to three conditions: 1. A region of interest (ROI) was defined. The area covering the possible locations of the laser light was manually selected. 2. The points must be green: the ∗ channel on the CIELAB image must be below a certain threshold, 0.1 ≤ ≤ 0.9 . Values for were defined differently for different consecutive series of images, due to inconsistent lighting throughout the tests. 3. The points must be significantly brighter than the background ( ) . The background was obtained by applying a Gaussian filter to the RGB image’s red channel, which made the laser light less visible, and the brightness was the CIELAB image’s ∗ channel, which emphasised the laser light. The ratio between the brightness and the background must be above a certain threshold, 0.8 ≤ 2 ≤ 1.7 , 2 also varied for different series of images based on the lighting conditions. Fig. 4 shows transformation evolution from RGB to binary images following the abovementioned steps. So, the transformed image is shown in Fig. 5 for Half_one and Half_two experiments. Laser points on both acquisitions have been extracted from the marks position (which were identified by the magnetic bases held onto the tower wall) and they were fused together to form the whole tower profile. The simplified and aligned point cloud of the tower, after merging the two halves, is shown in Fig. 6 from different views. The tower’s inner diameter has been measured as ≅ 4015.77 mm from 3D LSS experiment. Comparing to the reported value, ≅ 4177.00 mm, a deviation of 3.70 % has been achieved which is promising. This small difference on the obtained result strongly implied that the supporting methodology is accurate and robust in the field experiment application.

(a)

(b)