Light section is 3-D imaging modality which can be considered as a special case of structured light imaging. The most basic light section setup consists of a line laser, a camera with known intrinsic parameters and the object to be scanned. Based on triangulation and the distortion of the laser’s line on the object’s surface, the three dimensional shape of the object can be recovered.
Depending on the application one can choose between different approaches to light section which put different constraints on the setup. Some setups for example require a calibration pattern in the background of the scene, a robot arm or some other mechanical device to move the laser, or rotary plates for rotating the object.
The Spatial Relation between Laser and Camera
Independently of the chosen setup the spatial relation between the scene, the camera and the laser must be known or calculated. There exist multiple approaches, including systems in which the relative position between camera, laser and the scanned object is known by design, estimated before using the system or, like in the example discussed in this section, estimated online.
Consider a setup which shows a foreground object, the background consisting of two orthogonal planes and the lines projected by the laser light source. In this setup the relation between the camera and the scene can be computed from the known geometry of the planar calibration patterns visible in the scene. Furthermore, it is possible to calculate the position of the laser from the laser lines on the background planes.
Detection of the Laser Line and Laser Plane Estimation
Ideally the laser line can be detected automatically in the camera’s image with high accuracy. Like discussed in the article about triangulation, the reliability and quality of detection results can be improved by choosing the right camera and by using additional band-pass filters in order to get rid of unnecessary image information.
When the laser line has been detected, the next step is the estimation of the laser plane, which is defined by the rays emitted by the laser. For calculating a plane three linearly independent points on this plane are required.
In the setup shown in Figure 2 these three-dimensional coordinates can be computed from the pixel coordinates of the laser line on the background, the calibration pattern and the camera’s intrinsic parameters.
Computing the 3-D Points on the Object’s Surface
Once the laser plane is known and the pixels which belong to the laser line have been identified, the three dimensional coordinates of points on the object’s surface can be calculated.
For each pixel on the laser line, a ray starting from the camera to the laser line in the scene can be calculated from the camera’s intrinsic parameters. By intersecting this ray with the laser plane the 3-D coordinate of the point this pixel represents can be computed. By calculating these coordinates for all pixels along the laser line and repeating the whole process for multiple laser lines one can reconstruct the surface of the object from one perspective.
If the whole 3-D model of the object is required multiple surfaces with different object orientations must be calculated and registered into a single model.
Evaluation of this Imaging Modality
By using high-quality line lasers and high resolution cameras objects can be scanned with very high accuracy. However, this comes at a cost, as objects have to be scanned line by line. This can be achieved by moving the laser (robot arm, piezo mirror, …) or by moving the object itself (rotary plate, conveyor belt, …). This limitation also results in the fact that only static objects can be scanned, making this technology unsuitable for real time imaging.