Use case

Power² for photometric stereo

Basler’s ace and frame grabber

Customer: MSTVision
Location: Ginsheim-Gustavsburg, Germany
Date: 2021

What is the Photometric Stereo process all about?

The photometric stereo process makes it possible to distinguish an object’s surface condition from its texture (e.g., the imprint of a pharmaceutical package from braille). With a fixed camera position, photometric stereo captures several images which vary only in terms of the direction of illumination. Among other things, they make it possible to determine the surface curvature of an object. Some surface defects can only be detected in the curvature image.

Left: Set-up of FPGA-based photometric stereo Right: 1. Average curvature 2. X – derivation 3. Y – derivation 4. Texture

What is the challenge in fault detection?

In the photometric stereo approach, a single camera takes at least three images under different lighting conditions, which typically results in the calculation of four images. The calculation in the CPU is time-intensive and often it isn’t possible to meet the required cycle times during production. Even modern processors don’t have enough computational power to examine surfaces seamlessly.

Top: 4 input images and the image of a digital camera Bottom: Images of average curvature and the albedo image

The solution is a vision system with photometric stereo method

MSTVision uses a fast Basler ace Classic camera for image capture in its own photometric stereo system. It delivers 180 frames per second at 4 MP resolution. The company implements the entire calculation of the photometric stereo in FPGA of the frame grabber microEnable 5 marathon VCLx from Basler. The captured “raw images” are processed directly in the FPGA and, without an additional load on the CPU, the frame grabber provides the results of photometric stereo as four optional images. With VisualApplets, Basler also has a development environment that enables the FPGA to be enhanced with its own functionalities. In this process, the frame grabber handles the actuation of the segmented lighting – synchronously with the image capture.

The extensive parallelization in the FPGA makes it possible to process very high data rates with a low power consumption with each frame grabber. Although it has not been possible to implement the photometric stereo process with high data rates so far, this is now feasible thanks to the FPGA acceleration.

The advantages of the vision system with Photometric Stereo method

Photometric stereo with high data rates, suitable for the production environment
Image processing directly in the FPGA
No additional CPU load
Much more compact system architectures

Basler ace area scan camera
The ace camera is available with a diverse selection of sensors, resolutions, and speeds—all at a great value.
Frame grabbers
These boards provide robust, high-speed image acquisition and processing as well as signal control. The FPGA processors ensure real-time processing.
C mount lens
The fixed focal length lenses fit Basler cameras perfectly. We always offer the optimal lens for different vision requirements.