WO2011023832A1

WO2011023832A1 - Binary active lighting system

Info

Publication number: WO2011023832A1
Application number: PCT/ES2010/000144
Authority: WO
Inventors: Silvia SATORRES MARTÍNEZ; Juan GÓMEZ ORTEGA; Javier GÁMEZ GARCÍA; Alejandro SÁNCHEZ GARCÍA
Original assignee: Universidad De Jaen
Priority date: 2009-08-27
Filing date: 2010-03-25
Publication date: 2011-03-03
Also published as: ES2363284B1; ES2363284A1

Abstract

The invention relates to a binary active lighting device for transparent parts with planar or curved surfaces, comprising a light source and means for generating a binary pattern of alternating light and dark fringes. The invention is characterised in that the binary-pattern-generating means consist of a grid of an LCD panel and the light source has high luminance. The inclusion of an LCD grid in the system facilitates the processing of data and renders the system more dynamic in terms of fringe size, etc.

Description

BINARY ACTIVE LIGHTING SYSTEM D E S C R I P C I Ó N

FIELD OF THE INVENTION

The present invention is applied to the field of inspection of parts using computer vision systems. In particular, it applies to the inspection of transparent parts.

BACKGROUND OF THE INVENTION

For a satisfactory surface quality control in transparent objects, it is necessary to characterize what types of defects are interested in detecting and of course the geometry and material of the surface under inspection.

An example of transparent surfaces to be inspected would be the lenses of vehicle projectors, which are transparent and have a convex shape with some flat regions or very little curvature. These areas are located, essentially, in their parts seen, so that the level of quality of the finish, that is, the absence of defects, is extremely important.

The inspection of transparent parts is a complicated task, since the environment must be perfectly controlled so as not to have unwanted reflections. Using the appropriate lighting system it is possible to obtain an image without reflections in which the characterization of defects is possible.

The selection of a suitable lighting system for an application based on computer vision requires an experimental phase in which the type of lighting to be tested will depend on the characteristics of the application we are interested in solving. Most industrial vision systems relate to one of the following types of inspection:

1. Dimensional quality inspection.

2. Superficial quality inspection.

3. Component assembly inspection. 4. Inspection of operational quality (correct operation).

Considering this type of pieces to be inspected, it is interesting that the lighting system is diffuse to avoid as far as possible shadows and reflections that may hinder the process of extracting the defect in the image captured by the vision sensor.

As in the case of transparent surfaces, reflectors also have unwanted reflections problems but their automatic inspection has been studied in greater depth. A widely used lighting technique is structured lighting (Aluze, D., 2002; Puente León, F. & Kammel, S., 2006; Leopold, J. et al., 2003; Seulin, R., 2001; Seulin, R., 2002). Imperfections on a surface cause significant deviations in the reflection of light and this property is used to detect them. Structured lighting is of a binary type and is composed of a succession of strips without light intensity and with maximum intensity. With these conditions, a defect appears in the image captured as a set of white pixels on a dark background. In these systems, mechanical devices are used to move the strips, which makes them not very flexible.

Among the existing lighting systems in the market, those that in principle could be able to highlight defects on transparent surfaces would be: a. Diffuse backlight system (based on fluorescent or LED).

In this form of lighting, the sensor is directed directly to the light source, observing a completely uniform white surface. Any object interposed between the sensor and the light source produces a shadow that is detected by the sensor as a black shape on the white background. This type of lighting is usually used for the detection of impurities in transparent or translucent objects. However, the disadvantage of this system is that it is only possible to characterize the types of non-transparent defects such as blackheads and grease spots.

b. Low diffuse angle lighting systems (also called dark field). It consists of a direct light of high intensity that is made to strike with a very high angle with respect to the normal of the surface object of inspection. In this way, it is achieved that on surfaces that have crevices or elevations, they interfere with the light path, producing bright areas. A typical application in which it is used is to highlight imperfections and contours on flat surfaces. In this case, the disadvantage of this system is that if it is applied to transparent objects, with areas of less curvature, it only manages to highlight surface dirt such as dust spots.

Another referenced lighting technique (Dallas et al., 2003 and Rosati et al., 2008), which is used when characterizing defects in curved reflective surfaces, does not use structured lighting and a mirror system that allows the image to be concentrated. reflected in the vision sensor.

To detect transparent defects on surfaces that are also, using structured lighting techniques, it is essential that the CCD camera is saturated, and for this to occur, high exposure times are required. This would not allow its applicability at the industrial level, since the image acquisition time, and therefore the cycle time, would prevent the inspection process from being carried out within the defined times to obtain the established rates

OBJECT OF THE INVENTION

The invention aims to alleviate the technical problems mentioned in the previous section. To do this, it proposes a binary active lighting device for transparent parts that comprises a source of lighting and means capable of generating a binary pattern of alternating light and dark stripes, where the means capable of generating the binary pattern are a mesh of an LCD panel and the source of illumination is of a luminance between

100 and 50,000 lux. The device can also comprise a light diffuser and / or a thermal insulator (composed for example of a double glass separated by air) between the light source and the LCD mesh. The lighting source may be for example a plurality of LEDs or a plurality of fluorescent tubes. The mesh of an LCD panel (liquid crystal) consists essentially of:

Switching power supply

Liquid crystal panel (LCD), which can also incorporate thin film transistor (TFT) technology.

- Electronic card / s for processing and conditioning the control signals of the liquid panel.

- Digital data interface (for example: DVI, HDMI ...) or analog (e.g. VGA).

The device Ia invention incorporates a backlight system based on this LCD system, which is subsequently illuminated. Subsequently, the images captured by the computer vision system in charge of performing the inspection are processed in order to obtain an aspect image in which the defects are characterized. .

The fundamental advantage of this binary active lighting system compared to other systems is the high flexibility offered by an LCD as a light source, facilitating its application to automated defect characterization systems. The orientation, size of stripes, lag between stripes in consecutive images and even the possibility of configuring a uniform background make the same lighting system suitable for detecting transparent defects of various sizes, orientations and even non-transparent defects, if the Stripless system, acting as a conventional backlight system. In addition, with the proposed system, it is possible to easily configure, at the software level, the optimal strip pattern to be able to reveal defects of different size, orientation and typology. This is another advantage of the proposed system with respect to other previous systems, since to make this type of adjustment it was necessary to structurally modify the lighting device with the consequent economic cost.

The second advantage of this system is its light output, which allows it to reduce the detection and characterization times of defects.

Another advantage offered by this system is that it eliminates the need to use previous simulation systems that allow optimizing the number and size of strips with which the definitive lighting system must have, since, the system itself serves as an experimentation platform for the adjustment of said parameters. Finally, one last advantage of this lighting system is that it significantly reduces the computational cost necessary to process the images captured by the vision system.

BRIEF DESCRIPTION OF THE FIGURES

In order to help a better understanding of the characteristics of the invention according to a preferred example of practical implementation thereof, the following description of a set of drawings is attached, where the following is illustrated as an illustration:

Figure 1 is a diagram of the device of the invention and its placement with respect to the surface to be studied.

Figure 2 is a flow chart of the defect characterization process.

Figure 3 describes the physical principle of defect detection with structured light. DETAILED DESCRIPTION OF THE INVENTION

The present invention raises as a solution to the technical problems raised above using an LCD mesh (liquid crystal panel with a switched power supply, electronic card for processing and conditioning the signal and a digital data interface), to which Ie has placed a source of overlighting on its back.

This can be formed, for example, by high frequency fluorescents, or LEDs, to which a diffuser could subsequently be interposed between the LCD and the light source. The different elements of the device (9) can be seen in Figure 1. The light source (1) and the LCD mesh (3) can be accompanied by a diffuser and / or thermal insulator (2). The emitted light (4) becomes diffused light (5) thanks to the filter action. This becomes structured light (6) as it passes through the mesh (3). After passing through the object to be inspected (7) the light is collected by the computer vision camera (8). The lighting source must be of high luminance, in a lighting range between 100 and 50,000 lux and as homogeneous as possible to avoid specific concentrations. It is also interesting that it does not reach high temperatures so as not to damage the LDC mesh that is positioned next to it. One option would be the use of lighting based on high intensity LEDs, in which case, and because it is specific lighting, it would be advantageous to use a diffuser. Another option would be the use of conventional high luminance fluorescents which could also include a diffuser and a thermal insulator, since they reach high temperatures but are cheaper than other systems. In this, the possibility of including a transparent or diffuse thermal filter between both the mesh and the fluorescents must be considered. The filter may for example be a double glass separated by air. The diffuser is made of a non-opaque material that allows the received light to be homogenized. The operation principle of the invention is described in more detail below:

Figure 3 shows the physical principle of the detection of defects by structured light, that is, with light (5) and dark stripes (6), and the type of image that the vision sensor (1) would obtain. In the event that there was no defect (4) on the surface, the refraction of the light would not reach the unlit strip. On the contrary, if there is a defect (2, 3), the refraction of the light would reach the unlit strip and the defect would appear as a set of illuminated pixels in the black strip. In addition, so that the defect appears contrasted and the processing algorithms necessary to characterize it are simplified, it is convenient to saturate the vision sensor.

The invention proposes to use an illuminated LCD mesh. This represents a great advantage with respect to the previous systems, since the LCD system in cooperation with a high luminance light source, in the range of 100 to 50,000 lux, offers greater flexibility than mechanical systems, having no limitations on the time to set and move the stripe pattern or even remove it and place a white background. To ensure the detection of any defect in the sensor's field of vision, the strip pattern must scan the entire surface. For the types of non-transparent defects that might be interesting to detect as well (black dots and other spots), it would be interesting to have a backlight system without stripes. In our case, the invention provides a more efficient lighting system because it also offers the following advantages:

• Have a dynamic strip pattern configurable in orientation and size (especially width).

• Have the possibility of eliminating the strips for, lighting with light

white, detect opaque defects.

• Greatly reduce overexposure times by illuminating the object.

The proportion between the width of strips of different luminance allows detecting defects of different sizes. This ratio will be adjusted to highlight the smaller transparent defect that we are interested in inspecting.

The method used to achieve the characterization of transparent aesthetic defects using the proposed lighting system is shown in Figure 2. First, the binary lighting must sweep the field of vision of the camera, therefore the image capture alternates with the fringe movement. All captured images are composed of what is called an aspect image. In this image the background appears in an intermediate gray level and the defects in a very high gray level. Being so contrasted, the segmentation (division of the image into regions that have similar attributes or properties) does not require complex processing techniques and will only be applied to areas of the aspect image that do not have borders, defined as regions of interest (ROIs). Once segmented defects in the image should proceed to the acceptance or rejection of the object transparent depending on the dimensions of the defect and the area of the crystal in which it has been detected.

Obtaining the aspect image Define the binary dynamic lighting as a periodic square waveform that must be displaced N times so that the width of the white strip Tb covers the entire period.

Being

• Δ: Offset between stripes in two consecutive images

• a: Duty cycle,

The homogeneous background of the aspect image will be obtained by making the average (M) of the N outdated square waveforms Δ.

NEITHER

1 V- ¹

M (x) = -} P (x - n - Δ) n €

Neither

π = 0 to T

Δ = s and i <s <a - T

Since the dynamic lighting is achieved with a monitor, the minimum displacement between strips will be that corresponding to a pixel. The smaller the displacement between strips, the lower the level of average intensity in the aspect image will be giving rise to a lower contrast between the defects and the background; however, the homogeneity of the fund will increase. The vision system of the invention is also not subject to inaccuracies of movement of the mechanical system. The appearance images obtained with the monitor and displacements between unit strips are of a homogeneity in the background such that it allows to obviate the preprocessing of the image before performing the segmentation of defects, greatly reducing the computational load of the image processing. We also define: /> (*) = ^ P (X - Ti - T)

π = - oo

If we consider a single period we will have:

Expressing p (x) of the sign function we will have:

sgn (x) + 1 sgn (x - a - T) + 1 1

p (*) =

The homogeneity of the intermediate gray level is critical when segmenting defects in the aspect image, so that in order to obtain an aspect image with a homogeneous background, in which the defects appear contrasted, it is essential to consider the following parameters:

• Size of the white strip.

• Period of the fringes.

• Offset between stripes in consecutive images.

At most the offset can reach the size of the white strip. In this way the number of images necessary to compose the aspect image will be smaller but the background of it will not be uniform. The homogeneity of the image (h) can be obtained from the equation

(Seulin, R., 2002):

M X N Where M and N are the dimensions of the image, μ the level of average intensity in the image and f (x, y) the intensity level of the pixel x, y. The closer h is to 1, the greater the homogeneity of the image.

Therefore, with the proposed system, it is possible to easily configure, at the software level, the optimal strip pattern to be able to reveal defects of different size, orientation and typology. This is another advantage of the proposed system with respect to other previous systems, because for making this type of adjustment was necessary to structurally modify the lighting device.

In addition, the proposed system eliminates the need to use previous simulation systems that allow optimizing the number and size of strips that the definitive lighting system must have, since, the device of the invention serves as a calibration platform for the adjustment of these parameters. Calibration is performed depending on the type of defect to be found, since the size of the strips is directly proportional to that size.

Claims

1. Binary active lighting device for transparent parts comprising a lighting source (1) and means capable of generating a binary pattern of alternating light and dark stripes (3), characterized in that the means capable of generating the binary pattern are a mesh of an LCD panel and the source of illumination is of a luminance between 100 and 50,000 lux.

2. Device according to claim 1 characterized in that it further comprises a diffuser (2) of light between the light source (1) and the LCD mesh (3).

3. Device according to any of the preceding claims characterized in that it comprises a thermal insulator between the light source (1) and the LCD mesh.

4. Device according to claim 3 characterized in that the thermal insulator is composed of a double glass separated by air.

5. Device according to any of the preceding claims characterized in that the lighting source is a plurality of LEDs.

6. Device according to any of claims 1-4 characterized in that

The lighting source is a plurality of fluorescent tubes.