Patent Application
20160025591
This disclosure relates generally to an apparatus and method for assessing the quality of a reflector. In particular, the invention relates to an improved deflectometry system for measuring the slope of a reflective surface and predicting its performance as part of a solar-collecting heliostat field.
In Concentrating Solar Power (CSP) plants an array of heliostats reflect sunlight toward a receiver mounted atop a central tower and containing a working fluid. One type of receiver transfers incident radiant energy to the working fluid to produce output high-pressure, high-temperature steam, which can later be fed to a turbine for electrical power generation. Heliostats are generally mounted on the ground in an area facing or surrounding the tower. Each heliostat has a reflector: a rigid reflective surface such as a mirror that tracks the sun through the actuation of a heliostat drive mechanism about at least one axis. Sun-tracking involves orienting the reflector throughout the day so as to optimally redirect sunlight from the sun toward the receiver and maintain the desired temperature of the working fluid.
The performance of reflectors in CSP systems is highly dependent on their shape and curvature. Large variations from a desired surface profile (e.g. flat or curved)slope amongst many reflectors in a field can hinder the supply of requisite flux to the receiver, as the amount of light reflected from the heliostats may not comply with expected results garnered from simulations and calculations by plant operators. Adherence to an intended shape is therefore a concern during the reflector assembly process.
Conventional techniques for measuring the curvature and surface quality of a mirror include Deflectometry, Interferometry, Photogrammetry, and Video Scanning. In deflectometry a digital camera is used to take a photograph of an image reflected in a surface to-be-measured. The image is typically a repeated pattern, such as a checkerboard. Software is then employed to discern variances between the image reflected in the target reflector and the image reflected in a reference surface. Photogrammetry involves attaching retro-reflective tags or stickers to an object surface and taking multiple photos of said surface using at least one camera, followed by quantitative analysis of mapping between the photographs. Interferometry is the process of measuring differences in surface height along the plane of a reflector by superimposing reflected light waves from a Helium-Neon laser. In the video scanning method (such as VSHOT, the Video Scanning Hartmann Optical Test), a direct measurement of surface slope variance can be obtained by reflecting a laser off a reflector to the aperture of a digital camera. Of these methods, deflectometry is the most well suited for quality control in an industrial assembly line due to the speed at which measurements can be taken (only one photo from each camera is required), minimal RMS error (high resolution digital cameras can ascertain minute variances between images), and the comparatively lower cost of the system hardware (no lasers required).
Even amongst deflectometry solutions there exist drawbacks that hinder its use as a quality checking system in high-volume manufacturing. Determining the slope map of a reflective surface can be a timely calculation, and there is often no correlation made between the curvature of the reflector and its effects on in-field performance. Because the slope calculations are dependent on precise positioning of the camera and the light source with respect to the reflector, the physical system is sensitive to even small perturbations and disturbances common in an assembly environment, requiring well-surveyed, configured, and calibrated components. Finally, conventional deflectometer techniques do not possess the capability to unambiguously identify specific features in a reflected pattern if the pattern comprises repeated sections. There exists a need for a more robust deflectometry system that can quickly and accurately take measurements of reflector surfaces for quality control in an assembly process.
An improved quality assessment method and apparatus for the automated assembly of reflectors is described herein, wherein the quality assessment system comprises a deflectometer apparatus and image processing software that can determine the slope profile of the reflector and predict its performance in a concentrating solar field. By incorporating improved algorithms for calibrating the apparatus setup and identifying specific features in a reflected pattern, the proposed system advantageously increases the speed at which measurements can be taken while minimizing the susceptibility of the system to physical disturbances.
The deflectometer apparatus comprises two or more digital cameras having a high resolution and a narrow field of view, a target comprising a flat panel imprinted with a known pattern, a holding fixture upon which a reflector (to-be-measured) is mounted, a calibration surface having a known or unknown shape and slope profile, and a console having a user interface through which the apparatus can be operated and images and data can be viewed. The holding fixture can be a free-standing structure or an articulating robot. The image processing software takes as an input the images taken by the digital cameras of the known pattern reflected in the reflector and calculates a slope profile of the mirror, wherein the slope profile is a map of the angular direction of reflected light from multiple points on the surface. For example, the image processing software can compare the location of features in the reflected image with features in the image of a calibration surface. Deviations between the pattern as reflected in the reflector under scrutiny and the pattern as reflected in a calibration surface can used to determine a slope profile for the reflector being measured. Other means of determining the slope profile of a reflector can include direct measurements of incident light angles from a single or composite image, or the differential reconstruction of multiple images of the same surface taken in different positions. Differential reconstruction is the process of generating a slope profile of a two-dimensional surface from a gradient field; in the present application that gradient field is found by differentiating between first and second images of a reflected pattern, wherein each image captures the pattern in a different position relative to the camera. This slope profile can be utilized to predict the performance of the reflector in a solar field; the results of said prediction help determine whether the reflector meets predetermined metrics for quality assurance.
A method of assessing reflector quality can comprise the following three Phases: Commissioning, Calibration, and Data Collection/Analysis. In the Commissioning Phase, the digital cameras are positioned and zoomed, and focused to effectively and clearly capture the space in which the reflector will be located. Additionally the system geometry (e.g. the spatial dimensions of the reflector and its orientation relative to the target screen and features in the pattern) can be measured and adjusted.
In the Calibration Phase, a calibration surface is placed in the same plane within which the reflector-to-be-measured will be positioned and is mounted to the holding fixture. Images are then collected of the calibration surface by all cameras and the image processing software locates the features of the known pattern and ascribes coordinates to those locations in the image. Images of reflective surfaces measured by the deflectometry apparatus after the Calibration Phase are compared to the images of the calibration surface to establish deviation from a reference surface. Comparing reflective surfaces to a reference surface improves the speed by which the quality assessment system can analyze a large population of units, since the calibration step need only be performed once prior to measuring a plurality of reflectors. The present invention discloses multiple means for establishing this reference surface.
According to a first embodiment of the method for assessing reflector quality, the Calibration Phase comprises the steps of:
mounting a calibration surface having a known slope profile to a stationary holding fixture in a plane parallel to the target screen, wherein the calibration surface has a larger mirror area than the mirror area of the reflective surface to-be-measured,
overlaying a template having the same dimensions as the reflective surface to-be-measured onto the calibration surface,
taking images of the calibration surface using the digital cameras,
determining the coordinates of the corners of the template using the image processing software, and
locating the features of the known pattern as reflected in the calibration surface.
Because the calibration surface can be fabricated to as close to an ideal quality as possible at a size large enough to reflect the entirety of the target screen, the system according to the first embodiment does not require an actuating holding fixture with moving parts.
According to a second embodiment of the method for assessing reflector quality, the Calibration Phase comprises the steps of:
mounting a calibration surface having an unknown slope profile to a movable holding fixture in a plane parallel to the target screen, wherein the calibration surface has a larger mirror area than the mirror area of the reflective surface to-be-measured,
overlaying a template onto the calibration surface,
translating the holding fixture in the plane parallel to the target screen along a first axis and taking an image of the shifted calibration surface,
translating the holding fixture in the plan parallel to the target screen along a second axis and taking an image of the shifted calibration surface,
performing a differential reconstruction of the images,
determining the coordinates of the corners of the template, and
locating the features of the known pattern as reflected in the differentially reconstructed image of the target screen as reflected in the calibration surface.
The method of differential reconstruction can supply a reference image from a surface without knowing the curvature or slope profile ahead of time. The result is that the system according to the second embodiment does not require a calibration surface fabricated under strict tolerances and can be used with a plurality of calibration surfaces.
According to a third embodiment of the method for assessing reflector quality, the Calibration Phase comprises the steps of:
mounting a calibration surface having an known slope profile to a movable holding fixture in a plane parallel to the target screen, wherein the calibration surface has a substantially smaller mirror area than the mirror area of the reflective surface to-be-measured,
repeating, at least twice, the step of translating the holding fixture in the plane parallel to the target screen and taking an image of the shifted calibration surface,
stitching the plurality of images together into a composite, and
locating the features of the known pattern as reflected in the composite image of the target screen as reflected in the calibration surface.
Because a smaller calibration surface is easier to fabricate under rigid tolerances for slope profile and curvature, the system according to the third embodiment can lower production costs when compared to the first embodiment.
In the Data Collection/Analysis Phase a reflector to-be-measured is mounted to the holding fixture in the field of view of the cameras and opposite the target having the known pattern. Images of the reflector are then collected by all cameras and provided as inputs to the image processing software. The image processing software immediately operates to locate features of the known pattern in the image and compares the locations of the features to corresponding locations found in the image of the calibration surface taken during the Calibration Phase. The software then calculates the slope profile of the reflector. The software can also perform a ray-tracing calculation to predict the performance of the reflector as if it were placed in various positions within a concentrating solar field. The software can also provide the RMS slope error between the slope profile and a reference or desired profile. Any or all of the software outputs can be supplied to the display of a console having a user interface, wherein an operator can make an informed determination as to the quality of the reflector. Alternatively, quality determination of the reflector can be automated by prescribing a threshold by which the slope profile or other performance metric is deemed acceptable. The deflectometer system and method therefore can be applied as a quality check in an automated assembly line for heliostat reflectors.
According to a fourth embodiment, the method of assessing reflector quality comprises only a Commissioning Phase and a Data Collection Phase and obviates the need for a calibration surface or a Calibration Phase. This method comprises the following:
A Commissioning Phase comprising the steps of:
positioning and orienting the digital cameras to have a sufficient field of view to encompass the volume of a holding fixture and reflector,
approximating the geometry of the system, and
mounting the reflector-to-be-measured onto a movable holding fixture in a plane parallel to the target screen, and
a Data Collection Phase comprising the steps of:
translating the holding fixture in the plane parallel to the target screen along a first axis and taking an image of the shifted calibration surface,
translating the holding fixture in the plan parallel to the target screen along a second axis and taking an image of the shifted calibration surface,
performing a differential reconstruction of the images,
determining the coordinates of the corners of the reflective surface, and
locating the features of the known pattern as reflected in the differentially reconstructed image of the target screen as reflected in the reflective surface.
Because the method of differential image reconstruction is used on each measured reflector individually, the system according to the fourth embodiment does not require a calibration surface or a surface specially fabricated to have an ideal slope profile.
An improved quality assessment apparatus for the automated assembly of reflectors is further described herein with reference to
The step of commissioning the holding fixture and target screen to be parallel to each other can include the steps of taking a measurement of the spacing between the target screen and the holding fixture at multiple points along their perimeters. This can be done, for example, using measuring tape, by referring to makes premade in the floor, by using a rangefinder laser, or another suitable technique. For example, a 5-directional laser can be positioned in the middle of the target screen and fired at the calibration or reflective surface to-be-measured. The reflector can be aligned by actuating or shifting the holding fixture such that the laser hits a marked spot on the reflective surface and reflects back to the position of the laser in the target screen. Rotational alignment of the laser can be achieved using horizontal lasers back-reflected from the edges of the target screen.
The step of calibrating the cameras can include the steps of focusing the cameras, utilizing the zoom feature, and taking an image of the target screen with each camera and processing the images using the image processing software to ascertain image quality. The cameras can be calibrated using the target screen or an alternate target screen having a different pattern. The step of actuating the calibration surface or reflective surface to-be-measured can comprise the steps of actuating the holding fixture arm to move the mirror along one or more axes in the plane parallel to the target screen.
The following details the technique by which a differential reconstruction of images is conducted. The first step is to take a first image of the reflective surface using the cameras, wherein the reflective surface can be a calibration surface or a reflective surface to-be-measured. Next, the reflective surface is translated a predetermined distance along a first axis, the first axis comprising one of either a horizontal or vertical axis, at which point a second image is taken of the reflective or calibration surface by each camera. The reflective surface can be translated, for example, by actuating the robotic arm of the holding fixture. Next, the reflective surface is translated a predetermined distance along a second axis, the second axis comprising the other of the horizontal and vertical axis (the first and second axes being orthogonal to each other), at which point a third image is taken of the reflective or calibration surface by each camera. The image processing software then compares the first and second images and finds the differences between them. Differences can include, for example, deviation in the coordinates of pattern features, as measured on a per-pixel basis (see pattern detection and coordinate assignment procedures below). These differences define the derivative of the horizontal and vertical slope error with respect to the first axis. The image processing software then compares the first and third images and finds the difference between them. Similarly, these differences define the derivative of the horizontal and vertical slope error with respect to the second axis. These two derivative quantities represent a gradient field, one that happens to be non-integrable. The slope error in a given axis, and therefore the components comprising the slope map, can be found by integrating these partial derivatives by, for example, a Poisson solver.
The following details the technique by which the features and corners of the reflective surface or of the template overlaid on the calibration surface are identified by the image processing software. Note that, where possible, the following steps are parallelizable; if the image processing software can perform multiple tasks, detection methods, or coordinate assignments simultaneously it will do so. For a known pattern exhibiting a checkerboard design, the first step is to determine the location of all boundaries between checkers, as it is the boundaries themselves that are used as features for image comparative purposes. This is accomplished by estimating the dimensions (height and width) of a single rectangular checker using, for example a pattern-matching algorithm or a fast Fourier transform (FFT) phase correlation. Next the processing software utilizes a feature detector to extract the likely positions of the light and dark checkers in the image. The feature detector scans each pixel, and for each pixel it identifies the pixels that appear one checker space ahead and one checker space behind in both axes (X and Y, row and column). The “checker space” is the previously determined estimated dimension of the checker in the corresponding axial direction. If the scanned pixel is brighter (or darker) than the pixels in all four neighboring checkers in both axes, then it is a white (or black) checker. Now that the colors of the pixels are defined, the edges of a checker can be represented as the boundary where the pixels transition from one color to the other (from dark-to-light or light-to-dark). The edge detection method can comprise, for example, a Canny edge detection operator that distinguishes between edge and “non-edge” pixels. The processing software then applies the edge detection operator in both axes and, using a binary threshold, ascribes a value to every pixel based on its edge or non-edge status. Then the processor detects connected regions of pixels having the same values and groups them into similarly connected regions using a blob detection method to define the entire checkerboard.
After the locations of all checker corners in the image have been detected, the next step is to assign each corner point a coordinate; coordinates are first defined relative to the entire grid, and are then associated with real world positions. The straightforward method is to divide the image into checker-sized spaces and assign coordinates to the corners based on the checker dimensions (rounding when necessary). However, because the reflected image of the checkerboard grid may be greatly distorted depending on the slope of the reflective surface or the alignment of the target and/or holding fixture, this approach may not be sufficiently robust. Instead the processing software breaks the set of points (the grid corners) into smaller grid subsets and assigns coordinates to the points in each subset. Then the full grid is re-created by overlapping all of the subsets onto each other and consolidating overlapping points to match neighboring sub-grids. To calculate the world coordinates, the image processer first locates the region expected to contain the non-repeating feature and scans each checker (as defined by the edge pixels above) to find the checker exhibiting the darkest hue. With the position of the non-repeating feature (the “darkest checker” in the previous scan) now identified, the processor can shift the previously assigned grid coordinates to account for any error. Finally the grid coordinates are converted to world coordinates using the approximate geometry of the measured system, such as the dimensions of the reflective surface, the position of the reflective surface relative to the cameras, the boundaries of the holding fixture, or other defined distances and locations.
With corner locations and the checker board grid having now been ascribed coordinates, the last step to fully mapping the reflective surface in the image is to identify the boundaries of the mirror. This step is performed by identifying, using the image processing software, the transition between the checkers and a dark border. To create this contrast, the holding fixture can incorporate a flat dark background that extends beyond the edge of the mirror as seen by the cameras. The dark background can comprise black material such as a cloth or sheet attached to the holding fixture and set around the perimeter of the reflective surface. With this addition the image of the reflective surface comprises the distorted known pattern surrounded by a black border. To estimate the mirror boundaries, the processing software performs a blob-detection of the image to discern regions of the image made up of pixels exhibiting similar brightness and then eliminating the regions, or “blobs” that are smaller than a predetermined fractional threshold of the expected checker size. The software then calculates the convex hull of the remaining blobs, which involves drawing the smallest possible polygon around the entire set of points. This polygon is then fit to the expected rectangular shape of the reflective surface. The result is a set of coordinates that define the edges of the mirror and encompass the checkerboard pattern. As an alternative, if only part of the mirror is outlined by the black border, such as the case in which only part of the background is blocked by the holding fixture having the dark surface, only that corresponding part of the image is fit to the shape of the mirror.
The following details the technique by which the slope of the mirror is calculated from perturbations in locations of the known pattern features between the images of the calibration surface and a reference surface or reference image. In prior steps, the image processing software determines a mapping between pixels in the image and coordinates on the reflective surface and a mapping between the location of pixels in the reflected image and the coordinates, or physical location, of the reflected light ray onto the target screen having the known pattern. These two transformations can be combined to yield a mapping from the physical position of the reflective surface to the destination of reflected light on the known pattern. Calculating the destination of reflected light on the target is highly sensitive to the identification of the mirror boundary, and so the same transformation between reflective surface coordinates and pixels is used for the images taken from all cameras. This provides the system with additional robustness in the event that the mirror surface is perturbed or moved during the test. While such a disturbance does create error in the location on the mirror where the light deflections actually occur, the magnitude of the deflection values themselves are unaffected, and it is these magnitudes which have a much larger impact on the ultimate performance assessment of the reflector.
Next, the image processing software compares the destination of reflected light from the image of the measured reflective surface to the corresponding destinations from the image of the calibration surface and calculates the discrepancies between these destinations in real world-coordinates. These discrepancies and the measured system geometry are used to calculate the deflection, an indicator of the mirror “slope”, at a point on the reflective surface. Each digital camera of the automated deflectometry system supplies an individual image from which the deflection of light at each point on the reflective surface can be estimated. If the elements of the deflectometry system have been aligned properly such that their geometry is as expected, these estimates should be equivalent. If the images from the multiple cameras produce differing results, it is likely that the calibration surface and the target were properly positioned parallel to each other. The software can correct for being “out of plane” by perturbing the image processing model to adjust a “plane factor” until the output from the multiple cameras register similar values for the measured deflections. Alternatively it is also possible to conduct the deflectometry measurements using a subset of the total digital cameras, including only one camera.
A user interface displays the results of the aforementioned calculations to an operator. For example, the user interface can display a floating point value or measure signifying the quality of the reflector. This floating point value on the display can be color coded for ease of interpretation. The user interface can also display a 3D rendering of the reflector shape having exaggerated distortions based on the slope profile. The operator can use the data and the visual depictions to make informed decisions about the quality control process for reflective surface assembly. Alternatively, the process can be automated to compare the result of the ray-tracing calculation to a predetermined metric and provide an immediate assessment of the reflector quality.
Various combinations and/or sub-combinations of the specific features and aspects of the above embodiments may be made and still fall within the scope of the invention. Accordingly, it should be understood that various features and aspects of the disclosed embodiments may be combined with or substituted for one another in order to form varying modes of the disclosed invention. Further it is intended that the scope of the present invention herein disclosed by way of examples should not be limited by the particular disclosed embodiments described above.
The present application claims priority to and the benefit of U.S. Provisional Patent Application No. 62/027,746, filed on Jul. 22, 2014, the entire disclosure of which is incorporated herein by reference.
| Number | Date | Country | |
|---|---|---|---|
| 62027746 | Jul 2014 | US |
