Real-time camera-based distance gage can replace position-sensitive detectors

An optical solution with embedded computing for distance measurement eliminates errors associated with surface tilt and surface shape in manufacturing process control applications.
09 April 2009
Gil Abramovich and Kevin Harding

In surface treatments such as depositions, fine machining, and dispensing of a cover or protective layers, online real-time height measurement may be required to control process quality. Typically, the distance between a reference point on the processing head and the surface should be monitored at high frequency (10kHz) to provide effective machine feedback. Other requirements often include non-contact measurement, maximum optical resolution (minimal combined errors due to aberrations, diffraction, defocus, etc.), a sub-millimeter laser spot size, compact and lightweight packaging, and the ability to calculate measurements prior to the surface treatment process.

In manufacturing process control, position-sensitive detectors (PSDs) are commonly used for distance measurement. They produce an analog voltage output that is proportional to the location of the laser spot image on the detector.1, 2 However, PSD technology has several drawbacks, such as performance degradation in the presence of multiple reflections, stray light, variations in the beam shape and intensity distribution, and surface tilt3 (see Figure 1). In addition, the single-unit packaging of the PSD is geometrically incompatible with various in-process measurement applications, as evidenced by several product examples.4, 5 One alternative is to use a camera-based solution that offers physical and algorithmic flexibility advantages. However, this type of system is often not capable of real-time processing due to host PC constraints.

We developed a high-speed, camera-based distance gage that eliminates some of the challenges associated with PSDs.6 Analysis of geometrical and illumination efficiency constraints suggested that imaging the reflection of a laser spot is the best method, so our system includes a linescan camera with sufficient resolution of 20 microns or better. The lens allows appropriate lighting, dynamic range, field-of-view, and depth-of-field in order to accurately segment a laser spot in the image.

Figure 1. Measurement error using a position-sensitive detector in the presence of multiple reflections and surface tilt.

The imaging subsystem is oriented at a nominal 45 degrees to the surface, and the laser illumination subsystem is at 45 degrees to the surface from the opposite side of the tool (see Figure 2, left). Therefore, the image and object planes (where the laser spot moves) are parallel and centered at the camera axis. This arrangement prevents keystoning (spatial tilt-induced magnification variation) and simplifies the system's calibration. The entire system is designed to fit inside a 150mm hemisphere above the region of interest, with an overall mass of a few hundred grams. The design also incorporates a means to handle occlusions, and prevents mechanical interferences between the gage and the tool tip or the processed surface.

The prototype comprises a compact 1024-pixel line-scan camera that features a 6.5mm field-of-view and a 45g, 35mm focal length, 2/3-inch format c-mount lens. A 635nm laser diode is used as the light source. A significant part of the prototype is an embedded image acquisition and processing unit (see Figure 2, right) that uses a field programmable gate array (FPGA). The first optical prototype is shown in Figure 3. With this system, the laser spot moves in a linear manner that directly correlates with the distance of the imaging system from the surface, and it does not exhibit any movement when the surface is tilted.

We developed a standoff camera system for distance measurement that determines the distance between the processing tip and the surface at the point of process. Measurements are made in real time without errors associated with surface tilt and surface shape. Further work is required to create a standalone prototype and to finalize the design and assembly of the FPGA solution.

Figure 2. Left: Overview drawing of the system components. Right: Proposed embedded solution. FPGA: Field programmable gate array. D/A: Digital to analog converter. GigE: Gigabit Ethernet.

Figure 3. Optical table prototype.

Gil Abramovich, Kevin Harding
GE Global Research
Niskayuna, NY

Gil Abramovich is an optical scientist and a project leader at GE Global Research. His work involves optical, algorithm, and system development for homeland security, machine vision inspection, and medical applications.

Kevin Harding is a principal scientist and the optical metrology leader at GE Global Research. He is internationally recognized for his expertise in 3D measurement technology. He has published more than 100 technical papers, taught over 60 short courses, contributed sections to six books, and received over 25 patents.