Structured Light Sensor Calibration

Professor Fred DePiero

Structured Light Range Sensor

Range sensors are devices that measure 3-D Cartesian locations on surfaces. This kind of measurement is similar to a surveyor's collection of geographic data for a topological map. However, range sensors collect surface data via non-contact, optical means. Range sensors are also commonly connected to a computer to automate acquisition, and any subsequent processing and display.

Structured Light (SL) is a well-established method of optical ranging based on triangulation. Typically a laser is used to project light into a scene in a known direction. Given favorable reflectivity, the intersection of the laser with objects in the scene can be observed by a video camera.

The optical paths of structured light triangulation.

Cameras may be outfitted with optical filters that are matched to laser optical frequencies. This produces very distinct imagery of the laser illumination. Electronic shutters are helpful when objects are imaged as they move continuously past the sensor. SL sensors are capable of generating accurate and very clean Cartesian range data in real-time. These types of sensors can be built to operate on a wide range of standoffs.

The SL sensor uses a plane of laser light. The laser plane is generated by a cylindrical lens and illuminates a profile of range data in each camera image. The sensor has no moving optical components. The 3rd dimension associated with measurements is determined via the position of a mechanical slider used to move objects past the sensor (or alternatively, a conveyor belt).
Multiple cameras may be included to acquire color data, for example, or to reduce occulsions.

Photo of actual sensor components. The laser is a red diode type, 650 nm, and the laser plane extends in and out of the page. The sliding platform is driven by a stepper motor. The black and white camera is outfitted with an optical filter and is used to acquire the range data. The color camera acquires RGB intensity data along the laser profile.

Networked computing pipeline of 4 Linux boxes, used for acquisition, registration and visualization. Enclosure for PIC microcontroller, controlling stepper motor and laser.

Image of laser plane intersecting a cluttered scene, as seen by the camera of the SL sensor. This image was captured with NON-typical ambient lighting to reveal objects in the scene. Cartesian range measurements appear to the right, as a 'point cloud'.

The sensor captures images of the laser profile and computes Cartesian range data at frame rate. Range measurements of a machined plate have been collected with a standard deviation in height of 0.005 inches. Methods of calibration and acquisition are documented. Calibration images, analysis and models are available. Range scans are also available, including the individual images seen by the camera and the Cartesian range data.

The latest sensor includes a color camera.This permits both range and color data to be simultaneously acquired. In the range image to the right, the point cloud was resampled to form a dense image.

A SIPTool-based animation of the sensor acquisition process is available for download.

SL sensors can be used for a variety of shape or position measurements in industrial applications.They have been used to measure the flatness of circular saw blades, the radii of fillets, the depth of the punch in the top of a pop-open soda can, to name a few. Optical measurements such as these are sometimes required to measure the shape of deformable objects, such as washers or o-rings. Softer objects tend to deform when measured with micrometers, for example. Optical measurements do not apply deforming forces (ignoring photon pressure!)

An example of measuring the diameter of a washer follows.

Image from a black & white video camera, showing the intersection of the plane of laser light with a small washer. A binary version of the same image appears to the right, where the center of the laser contour is highlighted in red and the extent of the top surface of the washer is indicated by blue lines.

Thanks to a calibration procedure, it is possible to transform coordinates in the image domain to world coordinates. This was done for the two points on the circumference of the washer. Four such images were captured for a total of eight points on the washer circumference. A circle was fit to these points and appears below.

Range measurements, fit to the model of a circle and used to measure washer diameter.

Note the spacing of the chords across the washer is not uniform, to illustrate the point of uneven conveyor belt velocity.

A SIPTool-based demonstration of the washer measurement is available for download. The demonstration also uses Monte Carlo to simulate the measurement process, in order to estimate the accuracy of the washer diameter.

Professor Fred DePiero

Structured Light Range Sensor

More Structured Light Info and Links

This Structured Light Sensor has been developed by: