Method and device for pattern recognition.

0001The invention relates to a method and an arrangement for pattern recognition, in particular fingerprint identification, by placing the pattern on an image recording device for the optical processing of the pattern image to be recognized and processing and recognizing it via an electronic processing unit.

0002A fingerprint identification system is used, for example, as access control for security areas. DE 4220971Al and DE 3712089 A1 are known.

0003The technical solution is based on the principle of feature recognition based on the two-dimensional Fourier transformation.

0004The end sections, lines and number (features) of the fingerprint are used regularly.

0005So also according to DE 3712089Al, the identification of the papillary line pattern with a light-sensitive part generating a segmental image signal and a comparison of the image signal with preset signals.

0006EP 0308162 is also known for the method and arrangement for generating and evaluating fingerprint images. The optical system is characterized by the image recording device with lighting, prism, objective lens, correction prism, aperture to avoid aberration and row detector as a CCD element. There is a sample support detection to start the detection process, in particular by a pressure switch.

0007DE 3006677 describes a device for recording fingerprints using a display surface for recording the fingerprints and an associated evaluation station with an image transmission device, for example an optical imaging device with an associated television recording tube. For the dactyloscopic comparison, the conversion of the features of the fingerprint into alphanumeric characters which are compared in a comparison device with stored character strings is described, or an impression is compared optically with a stored comparison impression.

0008Also known is DE 4016832, a door locking device, in particular for motor vehicle doors, which has a sensor for recording at least one fingerprint and a memory for reference information and a comparison device which unlocks or locks a device when a match is detected.

0009However, all known representations and systems have significant shortcomings. So they are not able to distinguish a counterfeit impression (e.g. foil) from a living person. Furthermore, selected arrangements are very complex in order to master the optical problems during image recording, such as a uniformly sharp image on the sensor surface, aberration, and the pattern does not lie evenly or exactly on the defined contact surface.

0010The task was therefore to develop a system which overcomes the disadvantages described.

0011In addition, the identification should be done by recording the dislocation of clear subjective characteristics in order to enable an exact and actual identification of the living individual.

0012An arrangement for fault-tolerant pattern comparison, in particular through different positioning of the pattern on the image recording device, is to be created.

0013The object is achieved by the features characterized in claims 1 to 6.

0014The pattern recognition, in particular fingerprint identification, is carried out by placing the pattern on an image recording device for the optical processing of the pattern to be recognized. This image recording device consists of a triangular prism, one surface of which serves as a pattern support surface, a light source, an objective lens or lens system and an optical sensor, in particular a CCD sensor.

0015It is advantageous that, according to the invention, the pattern contact surface of the triangular prism is arranged at an angle alpha = 25 to 44 ° (FIG. 3) to the optical axis of the objective lens in order to prevent the system from being deceived by images or films placed on it.

0016Furthermore, the total reflection on the prism surface is disturbed by the application of a pattern, in particular a finger. It is particularly advantageous that the contrast properties result in a high-contrast pattern print image, primarily a fingerprint image.

0017According to FIG. 1, the arrangement for pattern recognition consists of an image recording device (1), an analog / digital converter (2), an image memory (3), an image memory controller (4), an image processing unit (5) and a live test (6).

0018The image recording device (1) consists of a triangular prism (7), objective lens (8), CCD sensor (9) and light source (10). (see Figure 3)

0019The light beams emitted by the light source (10) are coupled into the prism via the surface (F1), totally reflected on the surfaces (F2) and (F3) and projected onto the CCD sensor (9) via the objective lens (8). If a finger is placed on the prism surface (F3), the reflective properties of the prism surface (F3) change. This means that the total reflection on the prism surface (F3) is disturbed. The changed reflection properties create a high-contrast fingerprint image. The mountain sections of the fingerprint lines that touch the prism surface (F3) create dark lines. The valley sections of the fingerprint are shown by bright lines, the total reflection is not so strongly influenced at these points. Through an optimized arrangement of the optical components of the image recording device (1) according to FIG. 3rd a uniformly sharp image is generated on the CCD sensor (9).

0020The CCD sensor (9) is tilted in accordance with the Scheimpflug condition in its imaging plane (BE) to the optical axis (OA) (angle delta).<maths id="math0001"><math display="block"><mrow><mtext>tan (delta ') / tan (delta) = a / a'</mtext></mrow></math><img file="EP0640933A2_D0001.tif" /></maths>

0021Angle delta occurs between the optical axis (OA) and the auxiliary line (OE). The tilt creates a distorted image on the CCD sensor (9). The distortion can be described by means of a mathematical relationship and is eliminated by calculation. If the prism surface (F3) is at an angle alpha = 45 ° to the optical axis (OA) of the objective lens, the recognition system can be deceived with a fingerprint photograph. To prevent deception, the prism surface (F3) is arranged at an angle alpha = 25 to 44 °. If this angle is maintained, an inserted fingerprint photograph can no longer be recognized.

0022According to the invention, it is proposed to arrange the finger rest at an angle of 15 degrees to the base of the housing. This provides an advantageous rest.

0023The recording module advantageously consists of a prism framed by means of a thin metal plate, a prism holder, the support of the lens, the lens, an illumination device arranged for backlighting the prism, holding angles, the camera element and the lens.

0024It has proven to be particularly advantageous to use red light-emitting diodes for backlighting and to use a translucent, diffusing material as the material for the lens. According to the invention, the use of Makrolon, which scatters the color of the light sufficiently, is particularly advantageous.

0025An interference field is generated by means of the sheet metal plate surrounding the prism, which causes a field disturbance when the finger is placed on it, which leads to signal evaluation and is used for the living test. Via an infrared measuring device, which is advantageously designed as a component behind the support surface with a reflective body for reflecting the signal. A measuring device based on an optical sensor system is also proposed, which can determine the oxygen content of the blood from the movement of the blood and thus fulfills the living test outputs the result to the memory or control unit.

0026Figure 3 shows the optical structure schematically.

0027Figure 2 indicates the process flow.

0028The process scheme is shown in FIG.

0029The CCD sensor (9) continuously records images and sends a VIDEO SIGNAL to the A / D converter (2) and the image memory controller (4).

0030The A / D converter (2) converts the analog video signals into digital image information, so-called gray values, which are sent to the image memory (3).

0031The PICTURE-RECORD signal is set by the image processing unit (5), so that the WRITE-ENABLE signal is set by the image memory controller (4).

0032The image memory (3) can therefore store the gray values ​​received by the A / D converter (2).

0033If a full image is completely recorded and written into the image memory (3), the image memory controller (4) sends a PICTURE READY signal to the image processing unit (5).

0034The image processing unit (5) is thus informed that an image is available for evaluation and processing.

0035The gray value variance is checked by the image processing unit (5) in different areas of the recorded image.

0036If there is no finger on the prism, the variance of the gray values ​​is small.

0037When a finger is placed on the prism, the variance of the gray values ​​increases.

0038It is advantageous that only certain areas of the image are evaluated for determining the gray value variance, not the entire image. This avoids a high computing effort.

0039Images are recorded until the gray value variance exceeds a certain threshold value to be specified.

0040This is achieved when a finger rests on the prism and provides a high-contrast image.

0041This ensures that only images with sufficient contrast are processed further. The image processing unit (5) is not engaged in evaluating and processing images of insufficient quality.

0042The rejection rate is significantly reduced.

0043If the threshold value of the gray value variance was exceeded, the image processing unit (5) deletes the PICTURE RECORD signal. The image memory controller (4) clears the WRITE-ENABLE signal.

0044No new picture is written into the picture memory (3).

0045The captured image is now available for further processing.

0046The live test trigger signal is set by the image processing unit (5) and the live test unit (6) is caused to start. With the help of the living test, it is to be avoided that imitations, for example in the form of silicone dummies or severed fingers, can be used to deceive the system instead of a real living finger.

0047The captured image is binarized. This results in a drastic reduction in the amount of image information and thus a shorter computing time.

0048The binaryization takes place with a dynamic threshold.

0049The binarization threshold is not used for the entire image, because then quality losses occur in the binary image which are due to uneven lighting of the finger.

0050The image is divided into small quadrants, for each of which a separate binarization threshold is calculated.

0051It is particularly advantageous that quality fluctuations in the lighting have only a small influence on the quality of the binary image.

0052Another advantage is that parts of the image that have little contrast can also be binarized with high quality.

0053The minutiae (line ends, line branches, etc.) and the center point are determined in the binary image.

0054In order to identify the minutiae as such, their shape is evaluated.

0055For this purpose, only the edges, i.e. only the transitions between light and dark pixels, are required in the binary image.

0056It is advantageous that not all of the image information is required for determining the minutiae, but only a small part. This significantly reduces the computing effort. The image can be run through faster to find minutiae.

0057The positions of the identified minutiae and their species are recorded.

0058Another advantage is that not only positions or distances of the minutiae from one another but also information about their type for identifying the finger, which is described below, are used. This achieves a higher level of identification.

0059The binary image is distorted because the grayscale image was recorded via a prism. This harbors the risk that depending on the position of the finger on the prism with the same fingers, other relations between the minutiae occur.

0060In order to eliminate this danger, an equalization is carried out. Due to the high computing time, the equalization does not refer to the overall picture, but only to the coordinates of the determined minutiae. Thus, a finger can be rotated on the prism as desired, without this having an influence on the relations between the individual minutiae.

0061The live test unit (6) delivers the result of the live test.

0062If the live test is positive, the data obtained can be saved as a reference, depending on the operating modes, or compared with existing references.

0063In the "teach-in" operating mode, the data is saved as a reference on data carriers (eg hard disk, chip card).

0064A person to be trained must be authorized to store references in order to prevent undesired references from undermining the security of the system. For us, this is done using a master key, which only allows authorized persons to learn the system. The master key cannot be duplicated. The code on the master key is changed when it is used by a random generator and is also encrypted.

0065The type of encryption is also changed with each use.

0066In the "Compare" operating mode, the data determined from the fingerprint image can be compared with previously stored references. In the event of a positive comparison result, peripheral devices (e.g. door openers etc.) can be switched, or access to data (e.g. databases) can be permitted.

0067If the comparison result is negative, the test person is rejected.

0068After the comparison or learning process has ended or after a negative live test result, the image processing unit (5) sets the PICTURE RECORD signal and thus the WRITE-ENABLE signal by the image memory controller (4).

0069The image information digitized by the A / D converter (2) can now be written into the image memory (3) and fingerprint images can be recorded again.