WO2002071046A1

WO2002071046A1 - Method for optically enhancing contrast in high-throughput optical inspection

Info

Publication number: WO2002071046A1
Application number: PCT/IL2002/000176
Authority: WO
Inventors: Diana Shaphirov; Yehoshua Kulik
Original assignee: Camtek, Ltd.
Priority date: 2001-03-05
Filing date: 2002-03-05
Publication date: 2002-09-12

Abstract

In an automated optical inspection system, a method for enhancing contrast between a first material and a second material in a field of view on a workpiece comprises the following steps.a) Acquiring at least two different images;b) producing for each image a pair of light intensity distributions, a first of the distributions for pixel light intensity of light reflected by the first material and a second of the distributions for pixel light intensity of light reflected by the second material;c) processing each of the distributions to create a set of parameters;d) using the parameters to calculate a set of dynamic ranges by applying different weighting ratios;e) selecting the dynamic range from step d) which has the largest distance between the distributions; andapplying the weighting ratio from the dynamic range selected in step e) to images of subsequent scans of the workpiece and related workpieces.

Description

Method For Optically Enhancing Contrast In High-Throughput

Optical Inspection

Cross-Reference to Related Applications

This application claims the priority of US Provisional Application 60/272,780, filed March 5, 2001.

Field Of The Invention

The present invention relates to an method and system for obtaining optimal illumination settings in an automated optical inspection system for subsequent application to all image scans of related workpieces. More particularly, the present invention relates to a method for permitting an automated optical inspection system to perform an adaptive learning process whereby the reflective properties of different material zones on a workpiece are characterized based on a reference workpiece or frame and used to apply optimal illumination settings for subsequent workpieces.

Background Of The Invention

PCBs are comprised of an insulating substrate and a thin conductive layer clad to it. This application will primarily relate to the instance where the conductive layer is copper, however, it should be understood that this is for discussion purposes only and that any other conductive material which may be used in workpieces which are subject to automated optical inspection are encompassed herein. Also, this inventive method and apparatus apply to the inspection of substantially planar surfaces of other workpieces for the detection of surface flaws therein, such as wafers, and is not limited to PCBs (referred to hereinbelow as "WKP"). During a PCB's processing, the workpiece undergoes various surface treatments which often change the relative reflectivity properties of the laminate and conductive layers. This change in relative reflectivities is not necessarily uniform and creates complications in automatically distinguishing between laminate and conductor during the automated optical inspection of the workpieces.

Known automated inspection systems and methods are based on the difference between the optical reflectivity of conductive surfaces such as copper, and that of laminates. In the known systems, a workpiece is scanned, frame by frame, the scans are converted into digitized images and if alarms are received, then digital image enhancement processing may be automatically applied to those frames in order to further resolve the image such that the alarm is determinable to be false or requiring visual verification. However, this process has not been successful at reducing the number of total false alarms per workpiece to a satisfactory level. Additionally, it requires a relatively large amount of computer processing power in order to perform the digital enhancement of the scanned frames identified as having potential problems.

It has been recognized that manipulation of illumination conditions can be used to improve image quality. For example, it has been known that the kind of illumination of a PCB can reduce false alarms. Well known kinds of illumination which have been used include dark field, bright field, diffuse illumination (Lambertian or dome illumination), and oblique illumination.

However, none of these known methods reduces the number of false alarms requiring visual verification to a satisfactory degree.

The following patents are disclosed herewith and incorporated by reference herein insofar as they describe what was published in the art prior to the filing of the present application. US Patent No. 5,608,453 (hereinafter '453) to Gerber et al. and US Patent No. 5,455,870 (hereinafter '870) to Sepai et al. disclose automated optical inspection systems for the detection of flaws in microcomponents.

US Patent No. 6,207,946 (hereinafter '946) to Jusoh et al. and US

Patent No. 5,185,638 (hereinafter '638) to Conzola et al. disclose lighting systems for use with AOI systems such as those disclosed in the '453 and '870 patents.

Summary Of The Invention

The present invention provides a system for optically enhanced contrast for use in high throughput automated optical inspection by first determining the optimal illumination levels for at least two independent light beams incident upon a sample frame or workpiece and applying the same illumination level settings to all subsequent frames or workpieces of the same workpiece type prior to their scanning into digitized images. Once the system has optimized the illumination level for each light beam, each subsequently digitized frame or workpiece is then automatically inspected using known inspection processes. By using the method for optimizing the illuminations levels from the light beams as described hereinbelow in greater detail and then applying the optimized settings to the entire workbatch, the number of alarms which need to be visually verified is dramatically reduced.

The optimized illumination levels optimizes the contrast between zones of different materials, useful in the detection of surface features of PCBs, wafers and the like, and especially for the optical detection of surface flaws, thus overcoming difficulties and shortcomings encountered in the use of existing methods and systems.

In an exemplary embodiment of the present invention the system manipulates the illumination levels of at least two independent light sources through the use of an inventive computational method that analyses the reflected light intensity from a reference surface for the determination of the best illumination conditions conducive to the inspection of PCBs and other WKPs. This inventive method can be applied by using inventive software in some of the existing inspection devices or in special inventive devices.

This invention provides an automated method for the determination of the optimal intensities of light beams and an apparatus for its application by the acquisition and superposition either of [a] an image from two light beams on a single light sensor; [b] images from each of at least two light sensors in different positions being impinged by at least one light beam; or [c] images from each of at least two light sensors in the same position but each light sensor being sensitive to and impinged upon by light beams of different characteristics.

In one exemplary embodiment of the present invention, an automated optical inspecting system includes n light beams incident on a field of view, where n greater than or equal to 2, and where the intensity of each light beam is independently adjustable. A light beam is considered individual or independent of another light beam on the same field of view when the angle of incidence and the angle of coverage of the two or more light beams under consideration provide a maximum overlap that satisfies the following equation:

Where α and β are the incident angle and Δα and Δβ are equal to one half of the angle of coverage.

In the exemplary embodiment where two sensors in different positions are used to collect superposable images from a single light beam, then the sensors are considered as being in different positions where, in accordance to the abovementioned definition, the overlapped area is smaller then the sensor's acquisition angle.

Using the superposable images from either of the above-described exemplary embodiments, a processor employing the inventive algorithm digitally manipulates and arrives at an optimal light intensity for each light beam in the multiple light beam embodiment; or calculates an optimal image weighting for each of the digital image representations acquired by each sensor in the multiple sensor embodiments. Once the optimal light intensity or image weighting has been calculated for the reference frame or workpiece, the settings are imposed prior to all subsequent scanning and digitization of subsequent workpieces or frames.

In the exemplary embodiment utilizing multiple light beams, the intensities of the acquired beams vary according to the orientation of the emitted beam and the reflectivity characteristics of the surfaces (i.e. substrate or conductor). Thus, the objects of this invention are to provide a distinct difference between the substrate pixels intensity and the conductor pixels intensity, and, in addition, to eliminate as much as possible the intensity difference between pixels of the same kinds.

Brief Description Of The Drawings

FIG. 1 is a schematic view of an exemplary embodiment of apparatus which may be used in accordance with the present invention;

FIG. 2 is a schematic example of three beams incident upon nine pixels and their corresponding reflection;

Fig. 3 is a typical depiction of a copper - laminate two-peaked histogram and of its dynamic range;

FIG. 4 is a general flowchart of the inventive method;

FIG. 5 is a table of initializing parameters;

FIG. 6 is a flow chart for the conductor and substrate pixels determination from reference code data; FIG. 7 is a flow chart for the copper and laminate pixels determination from an image;

FIG. 8 is a flow chart of the testing of a sample frame;

FIG. 9 is a flow chart of the scanning the frame by separate illumination channels (without PGM);

FIG. 10 is a flow chart of the testing of each acquired frame;

FIG. 11a is a flow chart demonstrating the calculations required for determining the values of parameters of the weighting coefficient;

FIG. 11 is a flow chart of the calculation of dynamic range of the contrast between light intensity distributions;

FIG. 12 is a flow chart which describes the general method for applying the present invention to apparatus having multiple light beams and/or multiple sensors;

FIG. 13 is a schematic drawing illustrating another exemplary embodiment of the present invention;

FIG. 14 is a schematic drawing illustrating another exemplary embodiment of the present invention; and

FIG. 15 is a schematic drawing illustrating another exemplary embodiment of the present invention.

Detailed Description Of The Exemplary Embodiments

With reference to FIG. 1 , an automated method and apparatus for the determination of the light intensities emitted from a specular source and from at least one pair of oblique light sources symmetrically positioned about said specular light source, for the optimal illumination intensities for providing optical contrast enhancement of an inspected surface. While more than one pair of oblique light sources could be utilized, only one pair is depicted in the description of this invention; however, the method and apparatus shown could be readily modified to utilize any number of oblique light sources. Moreover, this method and apparatus could be readily modified to utilize only one light source and a mechanism for dividing the source beam into multiple beams having different orientation and intensity.

Referring now to FIG. 1 , depicting a schematic view of an apparatus 1 implementing this invention, apparatus 1 comprises of three blocks:

- Image generation block 100,

- Image processing block 200, - External interface block 300.

Block 100 comprises of an X-Y platen 142 controlled by an X-direction drive 144 and a Y-direction drive 146. A WKP 148 is mounted on top of platen 142 to form a top, flat surface 149 parallel to a plane formed by the X-Y directions and facing away from platen 142, to be inspected according to this inventive method and by this inventive apparatus. Adjustable intensity specular light source 154 emits light beam 155 reflected by beam splitter 170 towards WKP pixels 151 , and whose specularly reflected component is acquired by image acquisition unit 158 after passing again through beam splitter 170. Unit 158 comprises of an array 159 of a plurality of photoelectric or similar pixels 160 such as CCD's, said pixels are referred to hereinbelow as sensor pixels 160. Sensor pixels 160 are typically arrayed in a straight line, for sensing a corresponding linear array of WKP pixels 151. An optical component of unit 158 such as a lens 166 is made to focus some of the light reflected from WKP pixels 151 onto sensor pixels 160. A diaphragm 167 controls the light flux through lens 166 and is electrically controllable to permit its computerized control by other components of this inventive system. Each one of sensor pixels 160 generates an electrical signal that represents the light intensity incident to it and reflected to it from a corresponding pixel 151 , said electrical signals are defined by a selectable lower and upper limits, defining a selectable range of gray scale values, said gray scale is divided into a number of values selected to be between 0 to 255 in this exemplary embodiment, so that said electrical signals form a gray scale representation of surface 149. Due to non-uniformities in the illumination intensity of pixels 151 and differences in sensitivities in areas or lines of CCDs, a correction known as Pixel Gain Modification, abbreviated as PGM, may by applied to pixels 160 outputs during their processing. Right light source 152 and left light source 156 emitted beams 153 and 157, respectively, are handled similarly, and both incidents the same area of pixel 151.

A light beam is considered individual or independent of another light beam on the same field of view when the angle of incidence and the angle of coverage of the two or more light beams under consideration provide a maximum overlap that satisfies the following equation:

The light intensities emitted by light sources 152, 154, 156 are controlled by controllers 172, 174, 176, respectively, such as by changing their voltages, so that each one of said light sources may function, and be controllable, independently of the others. Filters or other means for controlling the intensities of said beams can also be provided. Filters are also known to enhance the contrast between copper and laminate zones and may be used for that purpose in this invention.

Under current technology, large WKP must be divided into smaller parts such as 147, called frames hereinbelow, the output from each frame 147 being analyzed separately. A frame of this exemplary embodiment is a rectangle of contiguous WKP pixels, one side of which equals the number of the contiguous, linearly arrayed sensor pixels 160, such as 4096 (4K), and the other side equals 2048 (2K), which could be the number of steps taken during a full frame scanning, the total number of pixels of a frame in this exemplary embodiment is 8M.

Image processing block 200 comprises a digitizing unit 208 in functional communication with sensor pixels 160 through lead 162 and with computing unit 210 by lead 249 for digitizing sensor 160 pixels electrical signals and for transmitting said digitized signals or their gray value representations to unit 210 for their storage and further processing.

Unit 210 is also in functional communication with a memory device 212 for the storage of sensor pixel data, and with database 220. Unit 210 is also functionally connected to an X-servo drive 144 and to a Y-seπ/o drive 146 of platen 142 by leads 244, 246, to apply positioning command signals to said drives for positioning platen 142.

Unit 210 is functionally connected as well to light source controllers 172, 174, 176 by leads 272, 274, 276 to apply command signals to said light sources, controlling the light intensities of beams 152, 154, 156, respectively.

External interface module 300 of this embodiment may also comprise communication network interface means 310 to a network (not shown), wherein means 310 is connected to unit 210 by lead 312, and local interface means 320 such as a computer display, a keyboard and other interface devices connected to unit 210 by leads generally numbered 322.

Communication means 310 may be used for a centralized processing of data acquired by a plurality of apparatuses according to this invention.

Electrical leads may be replaced by I R or by wireless communication devices.

This invention is better understood by following the steps taken during its use and the analysis of typical scanning results. A selected frame of the first WKP in a batch to be inspected is scanned five times under different illumination conditions. The first scan is for the purpose of doing coarse thresholding for setting up the system, in effect defining what constitutes a substrate pixel and what constitutes a conductor pixel. Afterwards, three scans, each using a different light beam, generate image data whose processing according to this invention determines the simultaneous illumination conditions by all three light sources of the fourth scan of the selected frame, and of subsequent scans of the rest of the first WKP and of other WKPs of the batch. Any sequence of illuminations of the first three scans as listed hereinbelow can be used. Then a fourth scan using light levels calculated from the first three scans is performed for fine-tuning of the aperture setting as follows:

- First condition: Specular beam, at a selectable intensity illumination, abbreviated spec or s. - Second condition: oblique beam from the left, at another selectable intensity illumination, abbreviated I.

- Third condition: oblique beam from the right, preferably at the same illumination intensity as the left beam, abbreviated r.

- Fourth condition: Combined; specular beam at one computed illumination intensity, and oblique beams from the left and the right at another computed illumination intensity, wherein the illumination intensities are computed according to this invention to enhance the contrast between different material zones on the WKP and to lower the contrast among those zones.

A semi-quantitative illustration of the reflection from different parts of a WKP is shown in FIG. 2. Correlation coefficients express on macroscopic level the reflectance properties of materials textures viewed by CCD at the specific resolution. The physical context of the presented notation is illustrated in Fig. 2. A 9-pixels-array collects light reflected from a surface with an arbitrary texture profile is shown in Fig. 2. Each separate channel provides its own intensity distribution normalized to same Ramp = 2. This corresponds to a histogram representation with the peak position at the value 2 and peak half-width from 1 to 2 or from 2 to 3. Switching on all channels together and consequent normalizing to the same Ramp - 2 (sum of all channels distributions divided by 3) results in a nearly uniform surface appearance with a sharp Histogram peak. For the given intensity distributions the calculated correlation coefficients are : p_{t r} = p_{t s} = p_rs = 0.5. The influence of surface texture appears as histogram widening which can be compensated for by correctly adjusted illumination.

On regular WKP there is huge number of pixels thus a statistical approach is needed to analyze and determine the optimal illumination condition.

Each illumination condition and scanning of a frame of WKP surface 149 by a single light source, such as a specular source according to the first illumination condition, at a selected light intensity, generates a large number of digitized conversions of light intensities from said PCB surface, one for each WKP pixel 151 , which are processed to yield a digital representation of the gray scale value of each pixel. It has been found that said gray scale readings resulting from a single source illumination tend to cluster about two mean values: one mean value corresponds to WKP copper pixels and the other value corresponds to laminate WKP pixels. A histogram of the number of WKP pixels vs. the gray scale values of copper or laminate pixels may be approximated by two Gaussian distributions (for example), each distribution having its mean μ^c(j) and μ^L(j) and its standard deviation σ^c(j) and σ^L(j), for copper and for laminate, respectively, where j is index of the beam orientation. A typical histogram is shown in FIG. 3. It should be clear that other statistically useful distributions may be employed and that later references to Gaussian distributions, or Gaussian approximations or Gaussian approximated distributions are intended to mean the same thing..

Changing the illumination intensity by some factor changes both the separation of the distributions' means and their standard deviations by said factor, as long as most of these values are bound by the low and the high limits of the selected gray scale. Such a change could be done electrically, by changing the light source intensity, optically, by changing an aperture, or computationally, by multiplying each pixel electrical output or gray value by any selectable factor.

Similarly, successive illumination and scanning according to the second and the third conditions by the remaining two beams such as the right and the left, provides two additional pairs of scanning histograms. Care must be taken not to raise the illumination intensity above a certain value as this may increase the number of saturated sensor pixels, possibly causing loss of data.

Data from the resulting histograms also leads to some material invariants, as shown in Table 1. The term material invariants means that these parameters are independent of the light intensity, yet, they are depending on scene configuration (light beam-board-CCD position geometry), board orientation and inserted filter. The peak values referred to are the respective gray scale values of the copper or the laminate peaks, i.e. the largest number of pixels having the same gray value, abbreviated hereinbelow CP/LP.

Table 1

A "ramp", designated R hereinbelow, is defined as the desired position of the right peak, i.e. the conductor peak. The right and left beams are selected in this exemplary embodiment to be of equal intensity, this relative intensity being designated as "1", the relative specular intensity will be designated as "k". Therefore: CP-left = CP-right = R/(k+2), CP-spec = R*[k/(k+2)]. (1 ) As seen in FIG. 3, the distance between the copper and the laminate peaks, designated in gray scale units, is Δ(k), Δ(k) being a function of k.

The dynamic range DR (k) is defined in this exemplary embodiment as the distance between the 99 percentile point of the laminate pixel gray value distribution (the right side of the laminate distribution) and the 1 percentile point of the copper pixels distribution (the left side of the copper distribution). Other percentile values may be selected and applied in the analysis. It is a purpose of the description of this invention to show how the dynamic range is increased by proper illumination, and how the analysis is leading to a determination of the best k for contrast enhancement purposes.

The general expression of the dynamic range is:

DR(k) = R^*[Δ(k) - n^c(k)^*σ^c(k) - n^L(k)*σ^L(k)]. (2)

Where: σ^c(k) and σ^L(k) are the standard deviation of the corresponding intensity deviations, and n^c (k) and n^L (k) are factors measuring the widths of the peaks foots at the hight C in sigma units, where C is the percentage of the peak height. It could be shown that

Δ(k) = R^*[1-(k*l-spec + l-left + l-right)]/(k+2) (3)

While the gray value distribution of the copper and the laminate pixels is approximated by a normal distributed in this embodiment, other distributions and other values of n^c(k) and n^L(k) could be used.

σ^c(k) and σ^L(k) calculation

It is known that the mean of the sum of three normally distributed variables of the same size populations is equal to: μ(x1+x2+x3) = μ1 + μ2 + μ3 (4)

While the standard deviation of the sum of these variables is equal to:

σ(x1+x2+x3) = -jσf +

+ σ] + 2p_nσ_xσ₂ + 2p_uσ_xσ_i + 2p₂₃σ₂σ₃ (5)

Where pι₂, p₁₃, p₂₃ are correlation coefficients between each pair of the random variables.

Substituting the designations of this embodiment into the former relationship for the case of a frame scanned with specular, left and right illumination consecutively, the resulting standard distributions are:

σ(l + r +s) = Jσf + σ) + σ + 2p,__sσ_sσ, + 2p_r__sσ_sσ_r + 2_Pl__rσ_rσ, (6)

Normalizing all copper peaks to the ramp value wherein

CP(mixture) = CPJeft = CP_spec = CP_right = Ramp (7)

and applying the usual ratio of left:spec:right=1 :k:1 , leads to:

σ(l) = (k+2)-¹σ(l) (8) σ(r) = (k+2)-¹σ(r) σ(s) = k ^~1σ(s) for the copper or the laminate values.

Substitution of the three source standard deviation values gives:

(9) σ '(k) = (k + 2)^"1 (_σ/' )² + k² [σ_s' )² + (σ_r' + 2kp __sσ_s'σ + 2kp __sσ σ_r' + 2p_r'__lσ_r'σ

Where i is either copper or laminate.

n^c(k) and n^L(k) calculation In order to estimate functional dependence n(k) the histogram peak is approximated by Gaussian distribution. From the equation f(x) = C, where C is the percentage of the peak height, and where x = n(k)^*σ(k), a formula for n(k) is derived

n^j(k) = - 2\n(const * σ'{k)) (10)

Where "const" includes normalization factor and C value, and i is either copper or laminate.

For conclusion, the proper ratio is calculated as a maximum of the DR(k) (2), where Δ(k) is calculated by (3), σ^c(k) and σ^L(k) by (9) and n^c(k) and n^L(k) by (10). For these formulas 18 parameters is needed: Copper and laminate peaks (μ) *3 (left, right, spec) = 6; Copper and laminate sigma (σ) *3 (left, right, spec) = 6; Copper and laminate correlation coefficients (l-s,l-r,r-s) = 6.

These data is measured from the three images received for the same frame scanned with left, right and spec illumination separately. The images contain information on same copper and laminate regions under varied illumination conditions. Considering copper and laminate regions as two sets of random variables allows direct, immediately from the images, measurements of required data. To calculate all those statistical parameters it is obligation to determine copper and laminate regions at the selected frame that scanned. An option to perform this procedure is with CAD reference, if one exists, but it is possible to do it without a reference with a threshold algorithm.

The abovementioned formulae are used in the subsequent statistical analysis that leads to the determination of the illumination conditions required for the highest contrast enhancement by generating the widest dynamic range. FIG. 4 is a general flowchart 1 of an exemplary embodiment of this invention, followed by more detailed flow charts of the steps taken during the inspection of WKP according to this invention. The major steps shown in FIG. 4 are: - Initializing the apparatus, step 2, further elaborated in FIG. 5, by loading and storing in a memory 212 of image processing block 200 various initial values. Table 20 of FIG. 5 lists those initial values;

- Checking whether a CAD exists for the PCB, 10. A CAD as used hereinbelow means a computerized representation of the nominal geometric features of the WKP or PCB, and especially of its copper surfaces. If a CAD exists, then:

- Go to step 3, further elaborated in FIG. 6, in order to find from it the frame's copper and laminate pixels to serve as a reference, then proceed to step 7. If a CAD does not exist, then:

- go to step 4, further elaborated in FIG. 7, in order to find the frame's copper and laminate pixels from an image of an accepted PCB, to serve as a reference, then proceed to step 7. An accepted PCB serving as a reference is sometimes called a "gold PCB". - In step 7, which is further elaborated in FIG. 9, scan the frame under three different illumination conditions and save the scanning results: beam from the right, beam from the left and specular illumination. Then proceed to step 1 1 , wherein materials and image parameters are calculated from the scanned data of step 7. - In step 9 the illumination ratio k between the specular and the oblique intensities is a vector parameter in which the copper and the laminate peak gray value separation Δ(k) and the dynamic range DR(k) are calculated, and wherein DR (k) maximal value is found. The value of k for which DR (k) is maximal is the specular to oblique illumination ratio to be used during the whole workpiece scanning and for other workpieces on that batch.

- In final stage 12 an implementation is performed in order to set the system for achieving this calculated k. A detailed description of the abovementioned steps follows.

FIG. 5, generally numbered 2 , lists in Table 20 some of the suggested default parameters to be stored in a memory and used in the computations of this inventive method.

Peak_Min= the lowest accepted value for copper peak in a single channel illumination. Peak_Max= the highest accepted value for copper peak in a single channel illumination. Right_Peak_Min= the lowest accepted value for copper peak in a three channel illumination. Right_Peak_Max= the highest accepted value for copper peak in a three channel illumination. S_Min =the max number of saturated pixels for a single source illumination.

S_Dif= the max number of saturated pixels for "irregular" panels,

S_Dif > S_Min. Ramp_ref= Targeted value of copper peak for three source illumination. Ramp_channel= Preferred value of copper peak for a single source illumination. (F#) opt = recommended F#.

Saturated pixels in an 8-bit system are pixels with gray value that equals 255, and F# is a parameter that indicates the aperture situation.

FIG. 6, generally numbered 3, is a block diagram presenting the steps taken for finding the copper and the laminate pixels from a reference, if one exists. In step 30, initial parameters are selected from a table with default parameters, according to the material type and the cleaning method that have been used. Then, in step 31 , one out of several frames is selected for the illumination calibration; step 31 may be repeated several times, as is necessary. Steps 32, 33, 5 and 34 form a sequence of a loop in which the frame is scanned and some illumination conditions are corrected, if necessary. In step 32 the frame is scanned and its raw pixel electrical readings are corrected for any illumination non-uniformities, with Pixel Gain Modification (PGM). The corrected readings are then processed, step 33 ,to generate a frame histogram of the number of pixels and their reflected light intensities. The result is then tested, step 5, as further elaborated in FIG. 8. If the scanning results of the frame are unacceptable, step 34, repeat steps 32, 33 and 5 with another aperture, determined in step 5 of the previous cycle. If they are acceptable then proceed to step 35, wherein registration is performed by linear shift, by rotation and by dilation or contraction of the measured frame image, to best fit it to the sample frame image. In step 36 a definition of each pixel is performed and in final step 37 a calculation and storage the number of copper and of laminate pixels Nc and NL, respectively.

FIG. 7, generally numbered 4, is a block diagram presenting the steps taken for finding the copper and the laminate pixels from a scanned image of an accepted PCB whose CAD does not exist. In step 40, initial parameters are set from a table of default parameters, according to the material type and the cleaning method used. Then, in step 41 , the location of the frame to be scanned within the inspected PCB is selected. The frame is scanned in step 42, the pixel outputs may be corrected to compensate for illumination non- uniformities by performing PGM procedure, as is known, and the corrected outputs are then processed, step 43, to generate a frame histogram of the number of pixels vs. their reflected gray values. The result is then compared to a sample frame, step 5, further elaborated in FIG. 8. If the scanning results of the frame are acceptable (step 44) then proceed to step 45, setting a main threshold to separate the copper from the laminate pixels, and proceed to step 46, to calculate and store the number of copper and of laminate pixels Nc and N , distinguished by using the abovementioned threshold.

FIG. 8, generally numbered 5, is a block diagram elaborating the testing of the sample frame, referred to hereinabove. The resulting histogram of the number of pixels reflecting a certain light intensity vs. the light intensity should display two distinct peaks, one for laminate, and the other for copper. The number of clustered pixels having peaks is checked, step 50, and their gray values are compared, step 51. If there is only one peak, or both peaks merge, the histogram is displayed, step 52. One peak indicates that either an error occurred or manual intervention is needed, step 53. In this case the procedure is terminated, step 54.

If two histogram peaks are identified in step 51, then check the right cluster of the histogram peak, e.g. higher intensity pixels, in step 55. If the maximum gray value of the right cluster is lower or equal to the preset maximum value and is higher or equal to the preset minimum value (check in step 57) then the frame is ok as a reference and the process is proceed from the point where procedure 5 have been operated. If the right peak value is higher then the preset maximum value (step 55), the generated frame is too bright, thus a routine that closes the aperture is activated (step 56) in order to decrease the light flux that incidents the sensors. After calculating and implementing the new aperture, the system leaves procedure 5 with the value "0" for the sample frame, step 59, (i.e. the frame isn't ok) which means that another scanning of this frame should be done with the new condition. The same process is done if the right cluster value is lower then the preset minimum value, which means the generated frame is too dark. If this is situation, a routine that opens the aperture is activated (58) in order to increase the amount of light that incidents the sensors. After calculating and implementing the new aperture, the system leaves procedure 5 with the value "0" for the sample frame (step 59), which means that another scanning of this frame should be done with the new condition.

FIG. 9, generally numbered 7, is a block diagram describing the main steps of a first preparatory stage of the application of this exemplary inventive method. In this first preparatory stage, the inspected frame is scanned three times while a different light source illuminating it during each scanning: the described scanning sequence is of scanning with left, right and specular source, respectively, although different sequences are possible. The scanned data is then processed and the processed results are used for the determination of the light sources illumination intensities for achieving the highest contrast between the copper and the laminate pixels, as shown hereinbelow.

The steps depicted in FIG. 9, block diagram 7, are divided into four stages: initializing stage 70, left illumination stage 71 , right illumination stage 72 and specular illumination stage 73. Initializing stage comprises of steps 701 through 704, for initializing the apparatus. Step 701 , in which X-Y platen 142 and an inspected WKP 148, such as a PCB, to be inspected that mounted thereon, are controlled by an X-direction drive 144 and a Y-direction drive 146, to bring a selectable inspected frame of WKP 148 to its initial position, is followed by step 702, in which default voltage, aperture and filter are selected from a default data table, and by step 703 which sets the selected aperture and filter. The final step 704 of initializing stage 70 can be considered as the first step of left illumination stage 71 , by turning on the left source at the required voltage.

In the left illumination stage 71 , the frame is scanned in step 711 , the number of pixels vs. pixel gray level histogram is generated in step 712, followed by step 8, further elaborated in FIG. 10, in which the channel frame results, i.e. single source illumination results, are tested and illumination parameters are changed, if necessary. Channel frame results are the pixel image illumination levels obtained by illuminating the tested WKP by only one light source. If the results are unacceptable for further contrast enhancement analysis, step 713, then steps 71 1 , 712, and 8 are repeated with the new aperture value calculated in step 8 of the previous cycle. If the results are acceptable, step 713, then turn off left source 156, turn on the right source 152 at its required voltage, step 714 and proceed to step 715, moving back platen 142 to the initial position of the scanned frame.

In the right illumination stage 72, the same sequence of stages are occur and after the results are acceptable proceed to steps 724 and 725, turn off right source 152, turn on specular source 154 and move back platen 142 to its initial position. Next stage 73 is repeating the same sequence of stages for the specular source, while after an acceptable results have been generated, whole stage 7 comes to its end by turning off specular source 154 and moving platen 142 to its initial position (steps 734, 735).

FIG. 10 is a block diagram 8 of the steps taken during each illuminating stage of stage 7, for testing the captured frame. The ratio S between the number of pixels reaching saturation and the total number of copper pixels is defined as the saturation level S and is calculated in step 80, and the copper and laminate means μ^c, μ^Lare calculated, step 81. S is compared to a saturation threshold ratio, which should be much lower than 1. If S is lower then the saturation threshold, step 82, then check in step 84 if the value of μ^c is higher than the Peak_min value. If it is not, then check if aperture 167 is fully opened, step 85. If aperture 167 is not fully opened, go to step 87, enlarge the open of the aperture and exit to continue the same loop of the same stage on stage 7. If the aperture is fully opened, or if the answer for step 84 is "yes" , exit, with the frame OK.

If S is higher then the saturation limit, step 82, then check in step 83 whether the values of μ^c and of μ^Lare below Peakjnin and whether S is lower than S_Dif. If they are, exit with a status OK of the frame. This is the situation for "irregular" panels (with high level of saturation and low μ^c), the reason for this special condition is inherent property of the material and, hence, cannot be compensated by increasing or decreasing the light intensity. If the answer for step 83 is "no", then proceed to step 86, to decrease the aperture open and exit to continue the same loop of the same stage on stage 7.

With reference to FIG. 11a, there is shown a general flow chart showing how all the parameters, which are needed in order to find the best ratio k as seen in Step 11 of FIG. 4, are calculated according to the following tables of formulas:

The means and standard deviations are calculated using the formulas in table 2, and the correlation coefficients are calculated according to formulas in table 3. Table 2 _μCo_PPer

Where : i = 1,..., Nc = Number of Copper _ pixels j = \,...,Nl = Number of Laminate _ pixels

Table 3

FIG. 11 , is a block diagram 9 describing the main steps of the computation of the dynamic range DR as function of k, wherein k is the ratio between the specular light intensity and the oblique ones. In step 91 k is initially set to 0. A computation of the dynamic range is conducted in a loop between step 92 and 97 wherein k is increased by a small increment such as 0.01 in each cycle of the loop, and the highest dynamic range is found for a particular k, step 98. In step 93, the distance between the means of the copper and the laminate pixels distribution is calculated according to the proposed relationship of 93, followed by step 94 in which the copper and the laminate standard deviations of their gray scale values as function of k are computed. The results of step 94 are substituted in the formulae of step 95, wherein a standard (Gaussian) approximated distribution for the copper and the laminate pixels gray level is assumed, and the width of the distribution curve is assumed to include all of the pixels whose gray value probability lies between 0.01 and 0.99. Therefore dynamic range is defined in this embodiment as the distance between the right (0.99) value of the laminate cumulative distribution, and the left (0.01) value of the copper cumulative distribution. Other values of n^c and n^L could be selected and used, replacing the "0.01" in stage 95 by some other value.

The dynamic range DR(k) is then calculated, step 96, and k is incremented by some small value, until k reaches its maximum accepted value. The selected k is the one that permits the highest contrast or dynamic range DR(k) between the laminate and the copper pixels.

Returning now to Fig. 4, final stage 12 is the implementing of the calculated ratio k by adjusting the aperture and the voltage of the light source in accordance to k.

With reference to FIG. 13, an exemplary embodiment of the present invention has two sensors are used to collect superposable images acquired from a single light beam. In the embodiment shown the sensors are located in different positions. Generally, the sensors are considered as being in different positions when, in accordance with the following definition, the overlapped area is smaller than the difference of the sensor's acquisition angles. A light beam is considered individual or independent of another light beam on the same field of view when the angle of incidence and the angle of coverage of the two or more light beams under consideration provide a maximum overlap that satisfies the following equation:

Using the superposable images from either of the above-described exemplary embodiments, a processor employing the inventive algorithm digitally manipulates and arrives at an optimal light intensity for each light beam in the multiple light beam embodiment; or calculates an optimal image weighting for each of the digital image representations acquired by each sensor in the multiple sensor embodiments. Once the optimal light intensity or image weighting has been calculated for the reference field of view or workpiece, the settings are imposed prior to all subsequent scanning and digitization of subsequent workpieces or frames.

With reference to FIG. 14, a light beam having at least two wavelengths, i.e. polychromatic light, is reflected from a field of view on a workpiece to be used for determining the image weighting for subsequent imaging operations. The reflected polychromatic light is then received by a sensor adapted to detect and discriminate between the different wavelengths of light reflected and to construct at least two or more images, each from a different wavelength of the reflected polychromatic light. In an alternate exemplary embodiment seen with reference to FIG. 15, more than one sensor can be used to detect and discriminate between different wavelengths of the reflected polychromatic light beam. Additionally, instead of using a single polychromatic light beam, it is possible to use at least two monochromatic light beams. Thus it should be clear that generally speaking it is an important consideration that two images be generated, that are different but the of the same field of view, be used to generate the light intensity distributions.

With reference to FIG. 11 , one can see the general flow of calculations needed for generating the dynamic range of the weighted ratios for multiples of sensors and/or light beams. Ideally, one should employ one of the known mathematical methods referenced hereinabove for determining a global maximum to apply for finding the dynamic range, rather than simply using the maximum value allowable.

The description hereinabove of exemplary embodiments are presented in order to enable a person of ordinary skill in the art to design, manufacture and utilize this invention. Various modifications and adaptations to the exemplary embodiment will be apparent to those skilled in the art, and different modifications may be applied to different embodiments. Therefore, It will be appreciated that the invention is not limited to what has been described hereinabove merely by way of example. Rather, the invention is limited solely by the claims, which follow this description.

Claims

I Claim:

1. In a system for automated optical inspection, a method for enhancing contrast between a first material and a second material in a field of view on a workpiece comprising the steps of: a) acquiring at least two different images of the same field of view; b) for each of said at least two different acquired images, the step of producing a pair of light intensity distributions, a first of said distributions for pixel light intensity of light reflected by said first material and a second of said distributions for pixel light intensity of light reflected by said second material; c) processing each of said distributions to create a set of parameters; d) using said parameters to calculate a set of dynamic ranges by applying different weighting ratios; e) selecting the dynamic range from Step d) which has the largest distance between said distributions; and f) applying the weighting ratio from the dynamic range selected in step e) to images of subsequent scans of said workpiece and related workpieces.

2. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 1 , wherein said different images are generated by applying at least two light beams on a single field of view and acquired by a single sensor.

3. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 2, wherein said selected weighting ratio is applied by changing the intensity of said light beams in accordance with said weighting ratio.

4. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 1 , wherein said different images are generated by applying a single light beam on a single field of view and having reflected light from said light beam acquired by at least two different sensors.

5. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 4, wherein said selected weighting ratio is applied by changing the relative weighting of said sensor-acquired images.

6. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 1 , wherein said different images are generated by applying a single polychromatic light beam to said field of view, and collecting said reflected light from said polychromatic light beam with at least one sensor which is able to sense and discriminate at least 2 wavelengths of light from said polychromatic light beam and to construct an image for each of said different sensed wavelengths.

7. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 6, wherein said selected weighting ratio is applied by changing the relative weighting of each of said different wavelength images.

8. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 1 , wherein said parameters are selected from the group consisting of: the mean of each distribution; the standard deviation of each distribution; and the correlation coefficients between distributions of said first material and distributions of said second second material.

9. A method for enhancing contrast between a first material and a second material in a field of view on a workpiece in a system for automated optical inspection in accordance with claim 1 , wherein said weighting ratio is selected from a range from -∞ to ∞.

10. A system for automated optical inspection having an illumination source and a reflected light sensor, said system comprising image generation means for generating at least two independent images of the same field of view, processing means for deriving and selecting an image weighting ratio from said independent images and optical contrast enhancement means for applying said weighting ratio to said image generation means.

11. A system for automated optical inspection in accordance with claim 10, wherein said image generation means further comprises at least two different light beams, at least one of said at least two light beams having a controllable intensity, and a reflected light sensor.

12. A system for automated optical inspection in accordance with claim 10, wherein said image generation means further comprises at least one light beam and at least two reflected light sensors.

13. A system for automated optical inspection in accordance with claim 10, wherein said image generation means further comprises at least one polychromatic light beam and at least one reflected light sensor capable of discerning at least two wavelengths of light from said polychromatic light beam.

14. A system for automated optical inspection in accordance with claim 10, wherein said means for generating at least two pairs of light reflection intensity distributions further comprises means for calculating and selecting an optimal weighting ratio for weighting said generated images.

15. A system for automated optical inspection in accordance with claim 10, wherein said means for applying said weighting ratio comprises means for adjusting the intensity level of said illumination source.

16. A system for automated optical inspection in accordance with claim 10, wherein said means for applying said weighting ratio comprises using the selected weighting ratio in the superpositioning of the generated images.

17. A system for optical inspection of planar workpieces, comprising at least at least two light beams illuminating a single field of view, wherein the intensity of at least one of said at least two light beams is independently controllable, and said system further comprising means for acquiring comprising means for defining and implementing optimal illumination intensities for each of said light beams, wherein said optimal illumination is found through inspection of one selected frame of the workpiece and the whole workpiece is further inspected with same said illumination conditions.

18. A method for enhancing the contrast automatically, comprising the following steps: a) presenting the contrast level mathematically as the "dynamic range" which vary as function of k, which is the ratio between at least two independent light beams intensity; b) scanning a selected frame from the inspected workpiece with only one light beam incidents the frame, and generating a set of statistical results from the data that derived from the scanning result; c) repeating step b for every single light beam; d) finding the optimal ratio (k) by calculating said dynamic range with said statistical results and finding its max value, and e) adjusting the system parameters in order to achieve said optimal k.