DLP face exposure 3D printer model data rectification

1. Overview

In the use of DLP as a light source for surface exposure 3D Printing system, we often encounter the projector or light machine projection of light, deformation, fuzzy situation. These situations can lead to the following problems: Print model surface has a pixel texture, not enough smooth model local exposure distribution uneven, thick and thin, in the printing of very fine parts especially when the light source trapezoidal deformation, resulting in the data will be biased lens refraction and the physical deviation occurred, Causes barrel distortion or pillow distortion in the single layer

These problems, when the accuracy of the printing model is not high, the exposure is not obvious, once the printing model to face the industrial production environment, these problems are very deadly. This paper gives a general approach to the above four problems. 2. Pixel pattern

The pixel pattern is caused by the limitation of the projection resolution. Typically, the distance between each pixel is 50~100um, which is the initial resolution value of the light source and also the smallest unit. The distance between the smallest units, after the model is formed, on the surface of the model to form a grain, very obvious. At present, the industry on this issue, there are several solutions to the idea, here Record: 1. Increase pixel accuracy in the xy direction

In general, our exposure is as follows:

How to further divide the existing pixels. The answer is to move the projection, moving the distance of half a pixel each time in xy direction. You can get the following effect:

By moving the entire exposure position, the pixel resolution of the model edge is added to achieve smoother results. 2. Introducing antialiasing models to edges

Anti-aliasing is common in many areas, where we use 3D printer technology to improve the surface smoothness of the model. For each edge of the model, we can extract it separately, with the edge of the exposure layer alone on the edge, according to the resin molding standard, anti-aliasing.

Extract Edge to data:

When edges are extracted, anti-aliasing is done separately for the edges:

3. trapezoidal deformation

The trapezoidal deformation is caused by the fact that the light source is not completely parallel to the projection surface. Trapezoidal deformation with manual to adjust, the error is very big, casually on the difference fraction a millimeter above, therefore, need to rely on the software according to external measurement results, to rectify.

The basic idea of implementation is as above. 4. Uneven light source

By dividing the first layer model into 40 regions, the intensity values of each region are measured, and each value of the local is adjusted according to the overall value. 5. Fish Eye Distortion

The projected lens, if not very good projection, is prone to fish eye distortion, as shown below:

The camera correction library provided by OPENCV is not available here because the library is going to import the original calibrated black-back image to get the raw input data, whereas in 3D printer system, the original data is a three-dimensional stereoscopic model which cannot be determined by the camera calibration parameters. Therefore, we should combine the model of print test to redesign the algorithm to realize the method of eliminating fisheye distortion.