NICOLAU: COMPACT UNIT FOR PHOTOMETRIC CHARACTERIZATION OF AUTOMOTIVE LIGHTING FROM NEAR-FIELD MEASUREMENTS

Transcription

1 NICOLAU: COMPACT UNIT FOR PHOTOMETRIC CHARACTERIZATION OF AUTOMOTIVE LIGHTING FROM NEAR-FIELD MEASUREMENTS S.Royo 1, M.J.Arranz 1, J.Arasa 1, M.Cattoen 2, T.Bosch 2 1 Center for Sensor, Instrumentation and System Development, Technical University of Catalunya Rambla Sant Nebridi 10 E08222 Terrassa, Spain 2 Laboratoire d Electronique de l ENSEEIHT, Institut National Polytechnique de Toulouse Rue Charles Camichel 2 BP Toulouse CEDEX 7 France ABSTRACT The present works depicts a measurement technique intended to enhance the characterization procedures of the photometric emissions of automotive headlamps, with potential applications to any light source emission, either automotive or non-automotive. A CCD array with a precisely characterized optical system is used for sampling the luminance field of the headlamp just a few centimetres in front of it, by combining deflectometric techniques (yielding the direction of the light beams) and photometric techniques (yielding the energy travelling in each direction). The CCD array scans the measurement plane using a self-developed mechanical unit and electronics, and then image-processing techniques are used for obtaining the photometric behaviour of the headlamp in any given plane, in particular in the plane and positions required by current normative, but also on the road, on traffic signs, etc. An overview of the construction of the system, of the considered principle of measurement, and of the main calibrations performed on the unit is presented. First results concerning relative measurements are presented compared both to reference data from a photometric tunnel and from a plane placed 5m away from the source. Preliminary results for the absolute photometric calibration of the system are also presented for different illumination beams (driving and passing beam). Keywords: Photometry, automotive, lighting, near-field, optical testing, headlamps, light sources 1. INTRODUCTION Headlamp testing is one of the most important problems involving new light source development, more when unconventional light sources and reflector geometries where no previous experience is available become dominant into the market. Though software modelling stays a good option, the experimental setups for testing the final real result involve long photometric tunnels, where, according to current normative, the headlamp is mounted onto a high precision goniometer and positioned 25m away from a high-accuracy photometer. Such installations are scarce across Europe and, given the performance of the headlamp is not good enough, it must be taken back to factory to introduce the desired corrections and then tested again. In addition, other relevant aspects related to headlamps, such as alignment issues, headlamp behaviour testing after a major repair, or the effect of ageing headlamps are not considered at all or considered in qualitative, imprecise ways along the lifetime of the light source. Technical inspection tasks, for instance, reduce to a qualitative alignment test, although car lighting is one of the key aspects involving the driver s safety. The interest of the technique becomes evident when one considers the number of patents filed in different countries with proposals to overcome the problem [1-3]. Among the solutions which have been proposed, the most simple is using a CCD-based photometer to record the final plane distributions (not avoiding the 25m tunnel problem, although). The most popular alternative to the goniometer is the multi-photometer, where the headlamp is placed close to the focal point of a big diameter Fresnel lens yielding an illumination pattern close to the far-field on a given screen. Some kind of multi-photometer array in that screen samples an illumination distribution similar to that of far-field. However, the positioning and alignment of the headlamp become critical under such conditions, as the headlamp is an extended luminance field, the object focal point is just a theoretical paraxial construction, and the aberrations of a single Fresnel lens are important. Users of such systems report the measurement becomes unpractical and the system results misused in many cases, despite the relevance of its results in, for instance, quality control in a production line. Less popular alternative methods involve photocell sensors positioned in planetary mounts in order to collect the light in all directions from the source [3], and other approaches.

2 The situation could be solved if a low-cost, simple and compact measurement unit was developed allowing headlamp testing at the manufacturer s installations without the need of a 25m-long tunnel, or at the mechanical workshop after repair. We call it a New Instrumental and COmpact Light Analysis Unit (NICOLAU). Bearing all this in mind, a combination of different optoelectronical techniques has been implemented allowing sampling the energy field and its direction of propagation in a plane close to the headlamp (typically at a distance of about 300mm). With this information, the far-field distribution in any given surface, in particular at the plane of interest for the normative, placed 25m away from the headlamp, is easily measured. Although the final part of the photometric calibration procedures is being undergone at present, the system has been already calibrated and absolute and relative energetic distributions compared to those measured using a photometer placed at a plane 5m away from the headlamp. In the future, the results will be compared to the ones obtained at a reference photometric tunnel. This paper has been divided in 5 Sections. Section 2 briefly describes the basics of the measurement principle involved. Section 3 is centred on the sensor hardware and the main factors considered regarding sensor performance. Section 4 shows the photometric calibration results and the photometric distributions obtained for different types of headlamps. Section 5 summarizes the main conclusions of the work and introduces the future directions of work of the sensor. 2. MEASUREMENT PRINCIPLE The goal of the measurement is a complete sampling of the luminance field at the defined measurement plane through deflectometric techniques which discretize the directions of propagation of the wavefront, photometric techniques which enable energetic measurements of the propagating energy, and image processing techniques allowing an efficient calculation of the final light distribution in the desired final plane. We will make a short look at each of them Deflectometric techniques Deflectometry provides a simple method for the characterization of the directions of propagation of one wavefront. Using a CCD camera with an objective lens focused at infinity, the pixels on the CCD array become slope measurers with a resolution dependent on the pixel size and effective focal length (EFL) of the lens [4]. Figure 1 shows the basics of the measurement technique under a 2-dimensional arrangement. Figure 1: Measurement of slope u (dx/dz). x is the pixel size along X axis. When the lens is pointing at infinity, all rays within a small slope range fall on the same pixel, and may be approximated as a single ray with the average slope value. This slope may be measured, as that ray will impinge in the same pixel that an auxiliary ray with the same slope crossing the lens through the object nodal point. The slope of the ray is thus defined by dx x u = = n dz EFL (1) dy y v = = m dz EFL where x (y) is the size of the pixel in the CCD array along X(Y) directions, and n(m) is the pixel number counted from the centre of the array along X (Y) axis. Thus, from a register at a given point

3 (x m,y m ) on the measurement plane, the directions of propagation of the energy at that point may be determined from the active pixels in the CCD array. Obviously, this implies a compromise on the value of EFL of the objective lens, which needs be achieved between the slope measurement resolution (higher with long EFL) against the slope measurement range (larger with shorter EFL) Photometric techniques To account for an effective sampling of the luminance field at each acquisition of the CCD camera, the energy value needs to be recorded. The basic idea is as simple as calibrating the CCD array for using each of its pixels as a photometer. The grey-level registered in each pixel in the CCD array will be used to evaluate the energetic value of the bundle of rays travelling in the direction (u,v) passing through the point (x m,y m ) of the measurement plane. This allows the complete evaluation of a five-element description (x m,y m,u,v,e) from each pixel of each image acquired in the measurement plane [5]. Obviously, the camera is displaced across the measurement plane for different (x m,y m ) sampling points to cover the complete spatial extension of the luminance field of the headlamp. However, photometry is not quite that simple when using CCD cameras [6], as may be proved by any scientific literature database search. The main concerns to be evaluated involve the spectral response of the camera, which needs to fit the photopic spectral response of the eye, the dynamic range of the CCD camera, the linearity of the response across the complete measurement range, the analysis of vignetting of the optics and filters, and the determination of a illumination against grey-level constant, among others. Section 3.2 and 4.2 sums up how some of these issues have been handled in NICOLAU Image processing Once the sensor has acquired a set of registers from the luminance field of the headlamp at near-field, displacing across the measurement plane at a number of sampling points (x m,y m ), the final distribution of the light source in the far-field needs to be computed at a number of points (x f,y f ) at the desired final plane. Image processing techniques, which mainly may be thought of as accumulation through projection techniques, are used to compute the final distribution. Fig.2 shows the complete NICOLAU arrangement. Figure 2: NICOLAU arrangement: sensor displaces across measurement plane (near-field) and imageprocessing is used to yield the final plane distribution (far-field). Due to the high amount of information collected by the sensor, a raytracing approach was considered to be too much time-consuming so a projection of the registered image onto the final plane was considered as much better. This approach involves simple summations of shifted images on a user-defined final plane grid. The measured image distribution at each point (x m,y m ) is projected onto the final plane with an spatial extension defined by the aperture of the sensor, yielding a superposition of extended measurement plane images whose energy values accumulate on the final plane grid. The final plane was fixed 25m away from the headlamp to reproduce the measuring conditions of the normative. Final computing time is reduced to only a few tenths of seconds, and could be performed simultaneously to image acquisition techniques.

4 3. SENSOR DEVELOPMENT 3.1. Hardware Though the measurement principle is quite simple, its implementation needed a collaboration of optical, mechanical, electronical and image processing techniques. The mechanical unit was fully developed for this project, involving a high-accuracy positioning system using DC motors and high-precision encoders over the wide lengths required for sampling the luminance field (600mm x 400mm), with repeatability and accuracy rates of 15µm in sensor position. The system is mainly a pair of sliders on high quality benches, with the associated electronics for motor control of the displacements, and the development of a high-end user interface for motor calibration and testing (fig.3a). The motor commandment is attained through the central NICOLAU user interface, placed in the PC unit which controls the complete measurement. The PC unit acquires the registers of illumination at the measurement plane through a conventional image-acquisition card. Scanning time is the main time-consuming procedure at present, dependent on the number of sampling points an accuracy demanded for the particular application considered. (a) (b) Fig.3: a) Experimental set-up with motor electronic modules: shutter control not shown ; b) Software interface: accumulation results window, showing the reference points of the normative In order to achieve good quality photometric measurements, a strategy for enhancing the dynamic range of an 8-bit camera was a must. The selected 2/3 camera allowed external shutter control through electronics, so a system for automatically adjusting the shutter through an external electronics unit depending on the amount of energy reaching the CCD was set up. The shutter control is also attached to the central PC unit (not shown in fig.3a), so once the image is acquired the optimum shutter range is evaluated automatically from the acquired image. The same PC holds the image processing algorithms which yield the far-field distribution in the final plane, including the position of the points used as reference in the normative (fig.3b). Driving-beam results for a commercial headlamp are shown in the display of the interface Sensor calibration Several issues were addressed along the calibration procedures of the unit; among them, we concentrate on distortion and vignetting of the optical system, and linearity and spectral responsivity of the sensor. Another key aspect is the sampling strategy at the measurement plane. Distortion needs to be considered as far as its effect implies an error in slope measurement with effects in the final reconstructed distribution. With the EFL defined to achieve a good dynamic range-resolution

5 compromise, a commercial objective lens with wide field (enough for a 2/3 CCD camera) and low distortion was selected. The error introduced in slope measurement by the distortion of the lens was around one pixel at edge of field (under %). No further appreciable distortion was introduced by the use of the different filters, as they were plano-parallel glass plates. Being not measurable due to it is low value, the theoretical distortion curve was included in the measurements for distortion compensation, though its effect is undetectable in practice. Another very relevant effect is vignetting, as the reduction in transmittance of the optical system from axis to edge of field needs to be compensated for making correct photometric measurements, including both the effect of the objective and of the filters and filter mount. In order to know the correction to be introduced, the sensor was exposed to a field of uniform luminance under normal working conditions. After acquiring measurements under different intensity values to account for repeatability in different luminance conditions, a transmittance curve for the main apertures of the objective (N=1.7, 2, 2.8 and 4) was obtained (figure 4a), so its behaviour could be curve-fitted and introduced in the software for compensation of the change of transmittance with the field of view. (a) (b) Fig.4: a) Transmittance of the optical system for different apertures from experimental measurements; b) Response of the system for the complete shutter range, showing its strong linearity Another key aspect to be verified was the spectral responsivity of the sensor, as far as a response close to the photopic response of the eye was mandatory to get reliable photometric measurements. The spectral response of the complete sensor was evaluated in 10nm intervals from 480 to 690nm by interchanging the camera with a calibrated radiometer to compare the measured radiance with the grey-level obtained in the sensor for each wavelength and for different radiances [7][8]. When computing the difference of the sensor spectral response with the photopic response of the eye, the difference yielded a f 1 value of 4.6%, not arriving to optimum photometric quality (f 1 <3%) but well above the limit of a good quality sensor (f 1 <8%) [9]. This value would be easily enhanced through a filter matched to the spectral response of the camera. The final verification regarding sensor performance involved testing the linearity of the response along the complete dynamic range, and the shutter law for photometric compensation of the energy values registered under different illumination conditions. This was especially relevant because of the dynamic adjustment of the shutter. In order to test this response, a LED was placed in front of the camera and its emission controlled through the circulating intensity. The sensor acquires the images at different intensities imposing no saturation occurs, and registers the average grey level obtained against shutter value, so the shutter law for compensation in computation and the linearity of the sensor are simultaneously validated. The system yields a strong linearity, as may be seen in fig.4b. Finally, an additional problem still needed to be considered because of the periodicity present in many of the modern headlamp designs. If the reflector yields a coarse periodic pattern (fig.5a), a periodic undersampling of some of the directions of the source occurs, turning into an aliased intensity pattern in the far-field distribution, with periodic dark areas in the final plane reconstruction. The reason is that at

6 near field some of the slopes are systematically unsampled so they do not appear in the final plane reconstruction. The solution has been achieved including a sampling strategy different than the normal square matrix of aligned rows-and-columns, by including a certain degree of interlacing between adjacent rows of the measurement plane sampling array (fig.5). This effect does not appear when regular, smooth reflector surfaces are used in the headlamp. (a) (b) (c) Fig.5: a) Example of headlamp with a periodic pattern on the reflector; b) Passing beam: aliased pattern caused by slope undersampling; c) Passing beam: correct pattern after optimised sampling strategy 4. RESULTS AND PHOTOMETRIC CALIBRATION 4.1 Relative distributions Once the sensor has been properly calibrated, relative distributions can be properly measured and compared before performing real photometric measurements, which will require a final photometric calibration step. These distributions may be compared with reference measurements performed at photometric tunnels. The comparison of distributions is performed after an alignment and rotation step of both patterns to centre and tilt them in equivalent positions and directions. Relative distributions have been measured and compared yielding very similar results (fig.6). It is stressed that the results are shown for passing beam in all cases, the more complex situation due to the very high dynamics of a beam which decreases strongly from the central intense fan to a dark superior area. Fig.6 shows the comparison of reference data, NICOLAU data, and the deviations of both plots, showing very good agreement for a headlamp different than the one presented in fig.5. Deviation values in both cases stay under the 6% value at all points, and 82% of the control points have error under 3%. It becomes evident the higher detail of the distribution measured with NICOLAU, which has a much larger number of measured data points, as may be seen from the level of detail of both plots Photometric calibration The final step for reliable photometric measurements is obtaining a correspondence between the accumulated grey-level measured and the amount of real illumination received. Once the linearity of the system has been tested, the problem reduces to the measurement of the constant which turns grey-level into lux, for different illumination conditions. Such constant is not be easily obtained from the measurements in reference tunnels, as far as such installations are programmed for testing a very limited number of control points. The measurements in fig.6a have been obtained from an especially designed test involving 77 data points, so small misalignments of the system cause important deviations.

7 (a) b) (c) Fig.6: All data in passing beam conditions; a) Reference data from measurement at photometric tunnel; b) Data measured using NICOLAU; c) Deviation of reference and NICOLAU contour plots A simpler system was designed for photometric calibration of the sensor, allowing a more dense sampling of the field. The near-field pattern was measured using NICOLAU and reconstructed at a final plane placed only 5m away, where the final distribution could be measured using a photometer. The correspondence of accumulated grey-level values and measured illumination values allows the determination of the proportionality constant which will allow photometric measurements to be performed, in a controlled environment. Preliminary results on such calibration are shown in fig.7, where data obtained from one headlamp under passing beam and driving beam conditions are compared when calculated by NICOLAU and when measured with a photometer. The similarity of the distributions is clear again, and even better then in the case of fig.6, due to the higher amount of available data points when measuring at a plane placed 5m away. Fig.8 shows the comparison of both normalised distributions, showing the small measure differences, and fig.9 shows the linear regressions values for both illumination conditions. The value of the proportionality constant is calculated for both distributions, yielding a conversion factor of lux/greylevel in passing beam conditions (r 2 =0.996), and lux/greylevel for driving beam conditions (r 2 =0.989). This means a deviation of 6% appears between both measurements, meaning such constant is equivalent in both graphs within the uncertainty of the experiment. Future measurements of other headlamps will confirm the proportionality value obtained and refine it. At last, such constant will become a fixed parameter of the system under given working conditions. 5. CONCLUSIONS AND WORK IN PROGRESS A system combining optical, electronical, and image processing technologies has been demonstrated which allows reliable measurements of far-field performance of light sources from near-field measurements. A CCD camera with a proper optical system, including a set of filters, scans a measurement plane placed at 300mm (aprox) from the headlamp. From the measurement of slopes and energy obtained, it is possible to accumulate the illumination values in the final plane, which has been selected as the 25m plane defined in the normative. The system implies the construction of a complete hardware unit and a careful calibration of many aspects of the sensor (from geometrical distortion in the optical system, to the sampling strategy in the final plane, not forgetting spectral characterization and linearity of the

8 (a) (b) (c) (d) Fig.7: Photometer against NICOLAU measurements for photometric calibration; normalized a) Passing beam, measured with photometer; b) Passing beam, measured with NICOLAU; c) Driving beam, with photometer; d) Driving beam, with NICOLAU sensor, among others). Results show a good performance of the sensor in the former aspects and a good comparison with available measurements from a reference photometric tunnel. Relative distributions measured at a distance of 5m with NICOLAU and with a photometer also yield very good results. Preliminary results for final photometric calibration are shown in order to convert the accumulated greylevel values to illumination values. Works are in progress in different directions. First of all, the absolute calibration values will be confirmed using different light fan types of different headlamps, in order to confirm the correctness of the calibration constant. Moreover, the system will be targeted to industrial applications, from alignment of headlamps in production lines or Technical Inspection Units to other light sources. In particular, the system will be used to validate different tail light units. The present system is only limited by the field of view of the sensor, so it may be applied to other types of light sources (LED, some luminaries) given the angular and spatial extension of the luminance field is not too large for the capabilities of the present sensor. In addition, the system has information enough to compute the light distribution in any given plane, allowing the simulation of the behaviour of the headlamps on the road or at given heights and distances, allowing to evaluate the visibility of road signals, for instance.

9 (a) (b) Fig.8: Absolute differences of the normalised distributions in fig.7: a) Passing beam; b) Driving beam. Notice the higher dynamics of plot (a) (a) (b) 4.0e+5 8.0e+5 3.5e+5 7.0e+5 3.0e+5 6.0e+5 Accumulative Value 2.5e+5 2.0e+5 1.5e+5 1.0e+5 Accumulative Value 5.0e+5 4.0e+5 3.0e+5 2.0e+5 5.0e+4 1.0e Illumination Value (lux) Illumination Value (lux) Fig.9:Linear regression values for determining the photometric calibration constant : a) Passing beam; b) Driving beam. REFERENCES 1. K.Ohana, Illuminance measurement of vehicle lamp, US Patent (1995) 2. T.Vernon, Measurement of the output of a lamp, GB Patent (1994) 3. W.Dabkiewickz, Emitted light measurer for lamps: has photocells on strip opposite lamp mount in inner frame pivoting in outer frame, DE Patent (1991) 4. J.Arasa, S.Royo, C.Pizarro Profilometry of toroidal surfaces with an improved Ronchi test, Appl.Opt (2000) 5. J.Arasa, S.Royo, J.Caum, M.J.Arranz Measurement of the photometric distribution of light sources by deflectometric techniques, Proc.IEEE 01TH (2001) 6. I. Lewin, R. Laird, J. Young, The application of video camera techniques to photometry, J.Illum.Eng.Soc. of North America, (1991) 7. F.Martínez-Verdú, J.Pujol, P.Capilla Calculation of the color matching functions of digital cameras from their complete spectral sensitivities, J.Im.Sci. and Technol (2002) 8. J.Y.Harderberg, H.Brettel, F.Schmitt, Spectral characterization of electronic cameras Proc SPIE (1998) 9. C.De Cusatis Handbook of Applied Photometry American Institute of Physics (1992)

10