Counting the amount of construction reinforcement using a laser scanner

General

This article provides an example of using laser scanner for the task of counting the amount of construction reinforcement. The task is to simulate the scanning process of construction reinforcement rods using a virtual scanner.

Description of scanned objects

The object is a piece of construction reinforcement lying on the surface of a table. A 3D model of the laid reinforcement has been generated. For simplicity, the reinforcement is presented as cylinders with a diameter of 20 mm.

Simulation

For the simulation, I use specialized software developed by me to model the scanning process, since often the transfer of real objects takes a long time or is impossible (if the object is too large or cannot be moved).

The scanner is located perpendicular to the ends of the construction reinforcement:

The scanner is fixed, while the reinforcement moves along the direction of the arrow:

The scanning results in the following point cloud:

In some places the end surface may not be scanned completely, as a geometric shadow is formed, but this will not affect the calculation of the quantity.

This occurs due to the geometric overlap of the reflected radiation by the adjacent rod (red rays).

Zoom in:

Point cloud processing

The point cloud during modeling was exported in .obj format. This point cloud and a video recording of the scanning process can be downloaded from the link: https://drive.google.com/drive/folders/1M7xOMPw1lK4YqvTAug78Xnc8nqWlbigr?usp=sharing

You can view the point cloud using the free Meshlab software. You can download this software from the link: https://www.meshlab.net/#download

Below is an example of processing to obtain the number of objects in a point cloud.

Original point cloud:

The side surface points are not required for the counting as they complicate the segmentation. It is recommended to filter the side surface points of the rods. For this, we use the SOR (Statistical Outlier Removal) method. Filtering result:

Next we perform segmentation into individual objects. In this case, 17 individual objects were obtained.

This is consistent with the original model, which had 17 rods.

Conclusion

1. Laser scanners are suitable for the task of counting the number of construction reinforcement rods.

2. When scanning, a part of the side surface will be scanned, which can be filtered out when processing the point cloud.

3. The ends of the reinforcement should be approximately aligned to minimize overlapping of one reinforcement by another.

4. Segmentation of the resulting point cloud into individual objects yields the number of construction reinforcement rods.