JP2008134710A - Optical information reader and optical information reading method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To extract information by using maximum a posteriori probability estimation, and to reduce computational complexity at that time. <P>SOLUTION: An input image photographed by an input device (CCD103) is stored in an input image buffer 171 of a memory 107 installed on a decode substrate 106. Data stored in an input image buffer 171, an extraction bar code image buffer 172, a pre-processing bar code image buffer 173, a M-scan line buffer 174, a high resolution scan line buffer 176, an edge emphasis high resolution scan line buffer 177 of the memory 107 are supplied to a bar code area extraction part 151, an image correction part 153 and a scan line acquisition part 154 of a pre-processing part 152, a high resolution execution part 155, a post-processing part 156, and a bar code reading part 157 of a CPU part 105. Furthermore, the identified characters are stored in a bar code reading result buffer 178 of the memory 107, and extracted to an output device 109. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an optical information reading apparatus and an optical information reading method suitable for use in, for example, a bar code reader. Specifically, it is possible to extract information that exceeds the resolution of the device using maximum posterior probability estimation, and further, the amount of calculation at that time can be reduced and processing in a CPU (Central Processing Unit) can be performed. Is something that can be done.

  In a wide range of fields such as distribution, logistics, postal mail, medical / chemical inspection, etc., as barcodes that represent information on parts with different light reflectivities as means for recognizing articles, documents, materials, specimens, and other various objects Two-dimensional code symbols are widely used. Conventionally, an optical information reader is used for a barcode scanner, a two-dimensional code scanner, or the like that reads such a barcode symbol or a two-dimensional code symbol.

  However, when reading information such as a bar code symbol or a two-dimensional code symbol with an optical information reader, the bar code symbol or the two-dimensional code symbol itself is contaminated, the bar code symbol or the two-dimensional code symbol, etc. It is inevitable that noise will be mixed in the read signal depending on the state of the paper described. Therefore, various techniques have been developed to remove such noise.

That is, for example, a technique for removing noise has been developed by using a combination of a high-pass filter and a low-pass filter (see, for example, Patent Document 1).
Alternatively, in some cases, a paper transfer noise is removed by realizing a transfer function using a Laplace operator or the like during filtering (see, for example, Patent Document 2).
Furthermore, studies have been conducted to increase the resolution of low-resolution images using maximum posterior probability (MAP) estimation (see, for example, Non-Patent Document 1).

Japanese Patent No. 2752870 JP 2005-293327 A Hardie et al., “Joint MAP Registration and High-Resolution Image Estimation Using a Sequence of Undersampled Images”, IEEE Trans. On Image Proc., VOL.6, NO.12, pp.1621-1633 Dec.1997

  Conventionally, as disclosed in Patent Documents 1 and 2, various techniques for accurately reading information have been developed in optical information readers. However, the purpose of the techniques disclosed in these Patent Documents 1 and 2 is to remove noise caused by contamination and the state of the display surface, and does not extract information that exceeds the resolution of the apparatus, for example.

  By the way, when reading a barcode using an imaging device, the amount of information that can be stored in the barcode is affected by the performance of the imaging device, the imaging position, and the like. Further, there arises a problem that the barcode cannot be read when the imaging position is far. In particular, this problem becomes significant when the number of pixels of the imaging device is small, but even when the number of pixels is large, in order to read more information at a time, it is necessary to accurately read the barcode from a distance. .

  On the other hand, in the prior art, it was considered that a thin barcode element could not be read unless it had a width of about 1 pixel on the image. On the other hand, like the technique disclosed in Non-Patent Document 1, it is conceivable to increase the resolution of an image photographed using an image processing technique. To do this, a huge amount of calculation is required, which is difficult to implement with ordinary CPU processing, or takes a long time.

  The present invention has been made in view of such problems, and an object of the present invention is to enable retrieval of information at a resolution higher than that of the apparatus using maximum a posteriori probability estimation, and This is to reduce the amount of calculation at that time so that processing can be performed even with a normal CPU.

  In order to solve the above-mentioned problems and achieve the object of the present invention, the invention described in claim 1 forms information according to the length of the bright part and / or dark part of the image, and also in the arrangement direction of the information. An optical information reading device for reading a binary image having an arbitrary width in a direction orthogonal to each other, an image scanning means for scanning a range of an arbitrary width a plurality of times with different phases, and a plurality of times An optical information reading apparatus comprising: information extracting means for extracting information from a signal read by scanning using maximum posterior probability estimation.

  The optical information reading device according to claim 2, wherein the image scanning means for scanning a plurality of times with different phases from each other comprises an imaging device arranged at a predetermined angle with respect to the housing. To do.

  In the optical information reading device according to claim 3, the image scanning means for scanning a plurality of times with different phases from each other is the same by shifting one pixel of the optical information with respect to an arbitrary scanning line imaged by the imaging device. And a means for taking out a scanning line equally divided between the scanning lines as a signal obtained by scanning a plurality of times.

  5. The optical information reading apparatus according to claim 4, wherein the information extracting means for extracting information using maximum posterior probability estimation is weighted at the transition between bright and dark portions of an image with respect to a signal read by a plurality of scans. And a calculation means for reducing a calculation amount by setting a weighting value in advance.

  Further, the invention described in claim 5 forms information by the length of the bright part and / or dark part of the image, and has an arbitrary width in the direction orthogonal to the information arrangement direction. An optical information reading method for reading a value image, wherein a range of an arbitrary width is scanned a plurality of times out of phase with each other, and a maximum posterior probability estimation is performed on a signal read by the plurality of scans An optical information reading method comprising the steps of:

  In the optical information reading method according to claim 6, a step of discriminating an arbitrary scanning line and a scanning line that is the same by shifting by one pixel of optical information with respect to a signal read by a plurality of scans; A step of extracting scanning lines equally divided between the determined scanning lines, and a step of extracting information from the extracted signals using maximum a posteriori probability estimation.

  In the optical information reading method according to claim 7, the maximum posterior probability estimation is performed by performing a matrix operation on a signal read by a plurality of scans using weighting at a transition between bright and dark portions of an image and weighting. The value is set in advance to reduce the amount of calculation.

  According to the optical information reading apparatus and the optical information reading method of the present invention, it is possible to extract information that exceeds the resolution of the apparatus by using the maximum posterior probability estimation. In addition, the amount of calculation at that time can be reduced and processing by the CPU can be made possible.

  The present invention relates to an arrangement of a sensor (CCD: Charge Coupled Device) in a barcode reading system and processing of signals obtained by the sensor. Bar code reading systems are used in various situations such as distribution and manufacturing. In addition, personal use such as information acquisition using a mobile phone camera is increasing, and here, all situations in which such bar codes can be used are assumed.

  A configuration example of the assumed system as a whole is shown in FIG. The system includes an element (CCD) for imaging a barcode, a scanner unit 100 including a CPU that processes the information and decodes (reads) the barcode information, and a terminal 200 to which the decoded information is transferred. Yes. However, when the CPU part which processes information exists in the terminal 200, or independently of the terminal 200 as shown in FIG. 1B, the decoded information is directly stored or processed in the scanner unit 100. In some cases.

  In such a barcode reading system, the present invention first relates to the arrangement of CCDs present in the scanner unit 100. That is, FIG. 2 shows a specific configuration of the scanner unit 100. Here, the light beam from the light emitting element 101 is irradiated onto the barcode 300, and the reflected light is imaged by the CCD 103 via the lens 102. In this case, by disposing the CCD 103 slightly inclined with respect to the housing 104, it is possible to acquire very useful information at the time of reading.

  The reason why the CCD 103 can be arranged slightly tilted in this way is that the same information is arranged in the vertical direction in the barcode 300. In other words, information on different positions of the barcode 300 can be obtained from the CCD 103 by tilting the CCD 103 with respect to the barcode 300 in which the same information is arranged in the vertical direction. Since the scanner unit 100 is held by a human hand, it is considered that the scanner unit 100 is often inclined even with respect to the barcode 300. It will be inclined.

  Information obtained by the CCD 103 is supplied to a decode board 106 including a CPU unit 105. That is, the method of processing the information obtained by the CCD 103 by the CPU unit 105 is also a main point of the present invention. This process increases the resolution of the information of the barcode 300, and enables decoding to be performed with higher accuracy. There are various types of barcodes, but the present invention is basically applicable to any existing type of barcode.

  FIG. 3 is a schematic block diagram showing the configuration of the decoding substrate 106 described above. In FIG. 3, an input image taken by the CCD 103 as an input device is stored in an input image buffer 171 of a memory 107 provided on the decode board 106.

  The input image stored in the input image buffer 171 is supplied to the barcode area extraction unit 151 of the CPU unit 105. As shown in FIG. 4A, the barcode area extraction unit 151, for example, binarizes an image (step S1), extracts edges (step S2), divides the image into blocks (step S3), and edges in the block Processing such as number counting (step S4), determination of a barcode existing area candidate (step S5), extraction of a barcode area by labeling (step S6), and the like are performed. The extracted bar code area image is stored in the extracted bar code image buffer 172 of the memory 107.

  The extracted barcode image stored in the extracted barcode image buffer 172 is supplied to the image correction unit 153 of the preprocessing unit 152 of the CPU unit 105. In the image correction unit 153, as shown in FIG. 4B, for example, luminance normalization (step S7), edge enhancement (step S8), and the like are performed, and the pre-processed barcode image buffer 173 in the memory 107 is stored. Remembered.

  The preprocessed barcode image stored in the preprocessed barcode image buffer 173 is supplied to the scan line acquisition unit 154 of the preprocessing unit 152 of the CPU unit 105. In the scan line acquisition unit 154, as shown in FIG. 4C, for example, scanning of the basic line (step S9), search of the scan line shifted by one pixel by the block matching method (step S10), and increasing the resolution by M times. In the case of performing, processing such as obtaining a scan line that equally divides the scan line shifted by 1 pixel from the basic scan line by M (step S11) is performed. The acquired total M scan lines are stored in the total M scan line buffers 174 in the memory 107.

  Further, for example, an inverse matrix of a matrix that expresses deviation / blur set in advance in the external storage device 108 is read into an inverse matrix buffer 175 of a matrix that expresses deviation / blur in the memory 107. Then, the signals stored in the M scan line buffers 174 in total are supplied to the high resolution execution unit 155 of the CPU unit 105, and, for example, as shown in FIG. S12), weight calculation (step S13), an inverse matrix of a matrix expressing shift / blur is input from the inverse matrix buffer 175 of the matrix expressing shift / blur, and the repetition is performed from the inverse matrix of the matrix expressing shift / blur. Processing such as calculation start (step S14) is performed.

  Further, the resolution enhancement execution unit 155 performs an inverse matrix calculation (step S15) by repetition using the inverse matrix auxiliary theorem, and determines whether there is a change in the weight calculation (step S16). When there is a change in step S16 (Yes), the process returns to step S14. Further, when there is no change in step S16 (No), processing such as high resolution scan line result acquisition (step S17) is performed. The acquired high resolution scan line is stored in the high resolution scan line buffer 176 of the memory 107.

  The high resolution scan line stored in the high resolution scan line buffer 176 is supplied to the post-processing unit 156 of the CPU unit 105. In the post-processing unit 156, as shown in E of FIG. 4, for example, processing such as edge enhancement (step S18) is performed. Then, the edge-enhanced and enhanced resolution scan line subjected to edge enhancement is stored in the edge-enhanced and enhanced resolution scan line buffer 177 of the memory 107.

  The edge-enhanced high-resolution scan line stored in the edge-enhanced high-resolution scan line buffer 177 is supplied to the barcode reading unit 157 of the CPU unit 105. In the barcode reading unit 157, as shown in FIG. 4F, for example, the divided scan lines are combined (step S19), the scan line is binarized (step S20), and the thick element and the thin element are identified (step S21). Element correction (step S22), character identification (step S23), and the like are performed. The identified character is stored in the barcode reading result buffer 178 of the memory 107. Further, the information stored in the barcode reading result buffer 178 is extracted to the output device 109.

  As a result, according to the above-described apparatus, as shown in FIG. 5, for example, a barcode image is taken as step 1, and in step 2, the barcode in the image is extracted as a barcode area extraction. The characteristics of the lens and the CCD are corrected by performing image correction, edge enhancement processing, and normalization in the processing. The parameters required for preprocessing are changed depending on the camera (CCD) used. Further, the resolution is increased as the procedure 4, the edge emphasis processing is performed as the post-processing of the procedure 5, and the result of the completion of all the processes in the decoding of the procedure 6 is decoded.

  When the barcode is photographed with the camera in the above-described procedure 1, the barcode is tilted and photographed. In order to take an image without tilting the user, the CCD itself is tilted with respect to the horizontal. Alternatively, the sensor unit is made movable so that a necessary inclination is automatically obtained when the user takes an image. This makes it possible to acquire very useful information at the time of reading.

  Further, in the higher resolution in the procedure 4, the higher resolution is performed by using a maximum a posteriori (MAP) estimation. At this time, the barcode is made into a block and the inverse matrix operation proposed below can be used to perform the operation at high speed. In addition, an appropriate weight is calculated. In the following, the increase in resolution in this procedure 4 will be described.

  FIG. 6A shows a conventional image enhancement technique using MAP estimation, and FIG. 6B shows a conventional image enhancement technique using MAP estimation according to the present invention.

  Here, in the conventional method of FIG. 6A, an image having an appropriate shift is input (step S31). Next, the initial value of the high resolution image is determined by interpolation (step S32), and the image shift is estimated (step S33). Then, the high-resolution image is estimated with the weight set constant throughout the image (step S34). Further, it is determined whether or not it has converged (step S35), and when it has not converged (No), the process returns to step S33. Moreover, when it has converged (Yes), it is complete | finished (step S36).

  On the other hand, in the method according to the present invention shown in FIG. 6B, a line whose deviation from the reference line is a multiple of (1 / magnification) is input (step S41). Next, a high-resolution image is estimated with the weights constant for the entire image (step S42), and an optimum weight is determined for each pixel (step S43). Then, a high resolution image is formed using the weight obtained in step S43 (step S44). Further, it is determined whether or not it has converged (step S45), and when it has not converged (No), the process returns to step S43. Moreover, when it has converged (Yes), it ends (step S46).

  Due to the difference in the above methods, in the conventional high image resolution using MAP estimation, it takes a long time to calculate the image shift and the high resolution image alternately and repeatedly. However, in the method of the present invention, it is possible to estimate the deviation of all images (lines) at first by taking advantage of the characteristic that the same data is arranged vertically in the barcode, and shooting at an inclination, High resolution can be achieved by using the minimum necessary image.

  That is, unlike a general image, since a barcode image often has a large difference in luminance, it is desirable to change the weight for each pixel. Therefore, the weight is changed depending on the image, and the high resolution image and the weight are repeatedly calculated alternately. By changing the weight, it is necessary to repeat processing as in the case of high resolution using conventional MAP estimation, but it is possible to calculate with a small amount of computation by devising the calculation method and speeding up the processing It is.

  Further, when the initial value is determined, the deviation is known in advance, and if the weight is constant as a whole, it is possible to perform in advance a matrix operation that takes time. As a result, in a short processing time, a better image is calculated than in the case of interpolation by the conventional method.

  In the following, the resolution enhancement using MAP estimation will be specifically described. Here, the method disclosed in Non-Patent Document 1 is used as a method using MAP estimation for increasing the resolution of a barcode image.

  High resolution using MAP estimation is a method of estimating the high resolution image z and the registration parameter s that maximize the posterior probability Pr (z, s | y) when the low resolution image y is given. . Pr (z, s | y) is expressed by the sum of logarithm logPr (z) of prior probability and logPr (y | z, s) of conditional probability by taking the logarithm and using Bayes' theorem. Thus, the problem of MAP estimation can be written as: (Note that in the following explanation, vectors are shown in bold in mathematical formulas, but are not shown in bold because they cannot be used in explanations.)

The above equation is controlled by two weights σ and λ. Assuming that the ratio of the two weights is w = σ ² / λ, if w is large, the luminance of adjacent pixels becomes a close value. However, since the barcode has a portion where the luminance changes abruptly from black to white, it is necessary to reduce the value of w there. In the proposed method, w = w1 when white continues (when luminance of 200 or more continues for 3 pixels or more) and black continues (when luminance of 50 or less continues for 3 pixels or more). In this case, the value of w was changed depending on the luminance such that w = w2. FIG. 7 shows the result when the weight is changed.

  Next, acquisition of a scan line will be described. In the one-dimensional barcode, data is arranged in one horizontal row, and the same data is arranged vertically. Therefore, it is only necessary to scan as one line horizontally. FIG. 8 shows an example A of a one-dimensional barcode and a state B in which it is scanned. Since the data obtained from the three scan lines shown in FIG. 6B are all the same, what is required as a result of increasing the resolution of the one-dimensional barcode is one-dimensional data in a horizontal row and two-dimensional data. There is no need to increase the resolution of the image.

  Therefore, according to the method of the present invention, one high-resolution one-dimensional data is estimated from a low-resolution barcode image using data of a plurality of scan lines. Even in the case of a two-dimensional code, the present invention can be implemented if a plurality of ranges where the arrangement of data in the horizontal direction of the elements is the same can be scanned.

In the following description, the resolution of an image is increased by a factor of four. In that case, in order to increase the resolution correctly, one pixel (one-fourth pixel at a lower resolution) in a state where the image is increased in resolution. Four lines shifted from each other are required. The scan line acquisition method is shown below.
Procedure 1: Shooting an image Shoot as shown in FIG. 9 so that a position shift occurs intentionally. This can be naturally realized by slightly tilting the camera sensor unit from the horizontal.

Procedure 2: Obtain a line shifted by 1 pixel Determine one basic line as shown in line (a) of Fig. 9, take the difference from the line shifted by 1 pixel from the other lines, and make the smallest difference. Find (line (b) in FIG. 9).
Procedure 3: Line acquisition Therefore, the line (c) that divides the lines (a) and (b) into four equal parts is a scan line that needs to be shifted by a quarter. Using these four lines (a) and (c), the resolution is increased.

  That is, when acquiring a scan line, instead of simply acquiring a single horizontal row of data, the average of three rows including the upper and lower rows is taken as a single scan line to achieve a higher recognition rate. did. FIGS. 10 and 11 show examples of the results obtained when the scan line is obtained from one row and the average of the three rows is used. It can be seen that there is a portion where the difference in luminance is clearly seen by using three rows, while the portion where the luminance difference is hardly seen in one row is used.

Furthermore, speeding up will be described. Here, we first consider blocking for speedup. The problem in increasing the resolution is the amount of calculation. The problems can be summarized into the following two.
(1) It must be a large matrix operation
(2) Inverse matrix operation

  Here, the problem (1) is first considered. In order to reduce the size of the matrix to be calculated, a large-size matrix operation may be replaced with a plurality of matrix operations having a small size by simply dividing the signal into blocks. However, if the signal is simply divided into blocks, the error at the divided boundary becomes large and becomes a problem. Therefore, here, this problem is avoided by dividing adjacent blocks so that they overlap. The block size was set to 60 pixels after high resolution experimentally. The overlap between blocks was 20 pixels.

Next, consider speeding up of the inverse matrix operation (2). The high-resolution image z is expressed by equation (3).
Here, an inverse matrix operation is required, and at least a 60 × 60 inverse matrix operation is required.

In this equation (3), C represents smoothness and is represented by equation (4).

Therefore, _assuming that W _s ^T W _s = A and d = [d ₁ , d ₂ ... D _N ], the inverse matrix in parentheses in equation (3) is expressed by equation (5).

Further, the inverse matrix of Equation (5) is expressed by Equation (6).
For the inverse matrix, it is necessary to perform the calculation of Equation (6) N times (60 times).

Here, W _i is the two kinds of 0.002 and 1.0. Therefore, W _i when adding 0.002 to the diagonal matrix A can be changed to two kinds of 0 and 0.998. On the other hand, since the ratio of W _i is approximately (0.002) :( 1.0) = 3: 1, adding 0.002 can reduce the calculation amount to about one-fourth. Furthermore, finally required data is a high resolution image ▲ z ▼ (^ on z) obtained by multiplying the inverse matrix obtained by the above method by W _s ^T y.

Therefore, when W _s ^T W _s = b is set and b is applied to both sides of Expression (6) from the right, Expression (7) is obtained.

Here, if A _N-1 ⁻¹ C _N can be calculated, it is possible to calculate (z) (^ on z) by Equation (7). In equation (6), the calculation is to obtain a 60 × 60 matrix, but if equation (7) is used, it will be a vector calculation and a reduction in the amount of calculation can be expected. A _N-1 ⁻¹ C _N is expressed as in equation (8) using equation (6).

Further, A _N-2 ⁻¹ C _N is similarly expressed by the formula (9).

By repeating this calculation until A ₀ ⁻¹ C _N is calculated, A _N ⁻¹ C _N b is calculated. Table 1 shows a comparison of the calculation amount of the 60 × 60 matrix in this case. Note that the calculation amount of the entire image is the calculation amount when the barcode width is 600 pixels.

  In Table 1, the amount of calculation is reduced to about one fifth by using the method of the present invention, compared with the conventional method using LU decomposition. In the past, 40,824,000 calculations were required to calculate the inverse matrix, but the number can be reduced to 2,952,000 using the method of the present invention.

Furthermore, the resolution was increased while changing the weight w2 using images obtained by shooting 9 types of barcodes from 200 mm to 25 mm apart by 300 mm. The experimental results are shown in Table 2.

  In Table 2, the result of changing the value of w2 shows that the optimum value of w2 differs depending on the distance. That is, a higher recognition rate can be realized by changing the value of w2 according to the distance. In this case, a high recognition rate can be achieved by using w2 of 0.05 when the shooting distance is up to 250 mm and 0.02 when the shooting distance is more than that. By changing w2 depending on the shooting distance and performing unsharp processing as post-processing, all 9 sheets up to 250mm were able to recognize 8 sheets of 275mm and 7 sheets of 300mm.

  In addition, the distance from 300mm to 10mm was taken and the limit of the distance that could be read was examined. The distance between the camera and the barcode is 215 mm in the conventional method, but 330 mm in the proposed method. When these are converted to the number of characters, the reading limit of the conventional method is 42 characters, and the proposed method is 63 characters, indicating that more information can be held. FIG. 12 shows the result of increasing the resolution using MAP estimation. From FIG. 12 (c), it can be seen that a better result is obtained as compared with that calculated by linear interpolation (b).

  Hereinafter, an embodiment in which the resolution is doubled will be described in detail with reference to FIG. The problem in FIG. 13 is that one line of high resolution (here, the uppermost reference line x) is estimated from the low resolution input.

Therefore, in the pre-processing in step 1,
(1) Normalize the whole input image in the range of 0 to 255 (here, it is in the range of 0 to 1 for convenience)
(2) Emphasize edges in the image (no emphasis is given here for convenience of explanation)
(3) A scan line shifted by one pixel is searched by comparing the red reference scan line at the input by one pixel with the subsequent scan lines.

here,
y0 shifted by 1 pixel => [0 0 0.5 0.9] (generated by folding the left edge)
y1⇒ [0.1 0 1.1 0.9]
y2⇒ [0 0 0.4 1]
And

  When these square errors are calculated, it can be seen that y2 corresponds to one pixel of y0. Therefore, y0 and y1 which is shifted by 1/2 pixel (1 pixel on the high-resolution image) are used to calculate high resolution.

Next, in the high resolution processing in step 2,
(1) Dividing the scan line (in fact, since the scan line has a length of 300 or more, it is divided into reasonable lengths, but in this example, it is not divided because it is sufficiently short).
(2) Interpolate input y0 to calculate weight ⇒ [0 0.25 0.5 0.75 1 1 1 1]
The smoothness weight for each pixel is determined by the interpolated information.
⇒ w = [w0 w1 w2 w3 w4 w5 w6 w7] = [0 0 0 0 1 11 1]

This weight is necessary so as not to impose a smoothness constraint at the position where the element switching occurs. Further, it is updated every time a high-resolution image is obtained by repetition.
(Here, for convenience, when the difference from the neighbor is large, it is 0 when it is small. However, when the luminance is 200 or more continues for 3 pixels or more, or the luminance 50 or less continues for 3 pixels or more. 1; 0.002 otherwise)

(3) If the high resolution information to be obtained is x = [x0 x1 x2 x3 x4 x5 x6 x7] ^T ,
What is necessary is just to obtain x so that becomes equal.

  However, since y0 and y1 have noise, correct results may not be obtained if the simultaneous equations are simply solved. Therefore, x is determined so that the sum of square errors is minimized.

These are written in a matrix
, And the because it is the sum of these ^{_{⇒x T (W 0 T W 0}} + W 1 T W 1) x-2 (W 0 T W 0 + W 1 T W 1) x + (y 0 T y 0 + y 1 T y ₁ )
It becomes.

Furthermore, considering the smoothness constraint condition (neighboring pixels take close values)
Must also become smaller, the problem when adding _{^{this, x T (W 0 T W}} 0 + W 1 T W 1) x-2 (W 0 T W 0 + W 1 T W 1) x + (y 0 T y 0 + Y ₁ ^T y ₁ ) to minimize x.

So, if you differentiate the above equation and set it to 0,
Can be solved.

If the inverse matrix is solved as it is, the amount of computation becomes a problem, so the inverse matrix theorem and
Is used.

A ₀ ^-1 depends only on how many times the resolution is increased, and can be calculated in advance. From these, the solution can be calculated recursively only by vector calculation, and it is not necessary to calculate at a position where the weight w _i is zero, which is very efficient.

The actual procedure is as follows.
a) As preparation, load A ₀ ^-1 into memory. B is calculated.
b) -1-1 Calculate A ₀ ^-1 C ₁ . A ₀ ^-1 C ₁ = [2 -6 6 -6 6 -6 6 -6]
b) -1-2 Calculate A ₁ ^-1 b. A ₁ ^-1 b = [0.1 -0.1 0.1 0.9 1.3 0.5 1.3 1.1]
b) -2-1 Calculate A ₀ ^-1 C ₂ . A ₀ ^-1 C ₂ = [-2 10 -14 14 -14 14 -14 14]
b) -2-2 Calculate A ₁ ^-1 C ₂ . A ₁ ^-1 C ₂ = [-2 10 -14 14 -14 14 -14 14]
b) -2-3 Calculate A ₂ ^-1 b. A ₂ ^-1 b = [0.1 -0.1 0.1 0.9 1.3 0.5 1.3 1.1]
b) -3-1 Calculate A ₁ ^-1 C ₃ .

Repeat as well
b) -3-3 Calculate up to A ₂ ^-1 C ₂ .
b) -3-4 Calculate A ₃ ^-1 b. A ₃ ^-1 b = [0.1 -0.1 0.1 0.9 1.3 0.5 1.3 1.1]
Repeat as well (actually, it is not necessary to calculate where w _i is zero)
b) -8-9 Calculate A ₈ ^-1 b.
A ₈ ^-1 b = [0.079 0.003 -0.085 1.167 0.950 0.956 1.038 1.103]

Therefore, the final solution for the set weight is
x = [0.079 0.003 -0.085 1.167 0.950 0.956 1.038 1.103]
It becomes. The number of steps in b) depends on the length of x.

(4) The weight is calculated for the result obtained in the above (3). If there is no change in the weight, the process ends and proceeds to the next procedure. If changed, the calculation of (3) is performed again with a new weight. In the example, since x = [0.079 0.003 -0.085 1.167 0.950 0.956 1.038 1.103] ^T , w = [1 1 0 1 1 11 1], which is different from the previous weight and returns to (3) again.

After the second iteration, the solution is A ₈ ⁻¹ b = [0.050 0.016 0.002 1.034 1.003 0.974 1.044 1.107] ^T
Since the weight is not changed from the previous time, this is the final solution.

  Further, in the post-processing of the procedure 3, the edge is emphasized, and in the decoding of the barcode in the procedure 4, the barcode is read. However, since these are the same as those in the conventional technique, the description is omitted.

  Thus, according to the present invention, it has been shown that the barcode can be read from a long distance by increasing the resolution of the barcode image using MAP estimation, and the amount of information that can be stored can be increased.

  In the conventional technique, the narrow element width is limited to about 1.0 pixel, but by using the technique of the present invention, it is possible to correctly read up to about 0.66 pixel. Assuming that a camera with a resolution of 640 × 480 is used, it is possible to read up to 63 characters, although 42 characters was the limit in the conventional method. Also, in order to read a barcode containing 29 characters, the conventional method required a resolution with a width of 420 pixels at the minimum, but using the proposed method, it can be read with a width of 285 pixels.

  As a result, it is possible to reduce the size or cost of the sensor unit of the barcode reader. In addition, the user has a high degree of freedom regarding the position of the sensor and the barcode when reading the barcode, and a more convenient barcode reading apparatus can be realized.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the claims.

It is a schematic diagram which shows the structure of the whole system to which this invention is applied. 2 is a diagram showing a specific configuration of the scanner unit 100. FIG. 3 is a schematic block diagram showing a configuration of the decode substrate 106. FIG. It is a flowchart figure for the description. It is a flowchart figure for the description. It is a flowchart figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description. It is a figure for the description.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 100 ... Scanner part, 200 ... Terminal, 300 ... Bar code, 101 ... Light emitting element, 102 ... Lens, 103 ... CCD, 104 ... Housing, 105 ... CPU part, 106 ... Decoding board, 107 ... Memory, 108 ... External storage Device 109 ... output device

Claims (7)

An optical information reader for reading a binary image having an arbitrary width in a direction orthogonal to the arrangement direction of the information while forming information by the length of a bright part and / or dark part of the image. And
Image scanning means for scanning the range of the arbitrary width a plurality of times out of phase with each other;
Information extracting means for extracting information using a maximum posterior probability estimation for the signal read in the plurality of scans;
An optical information reading apparatus comprising: The optical information reader according to claim 1, wherein the image scanning unit that scans a plurality of times with different phases from each other includes an imaging device disposed at a predetermined angle with respect to the housing. The image scanning unit that scans a plurality of times with different phases from each other determines a scanning line that is the same by shifting by one pixel of the optical information with respect to an arbitrary scanning line captured by the imaging device, and the like between them The optical information reader according to claim 2, further comprising means for taking out the divided scanning lines as a signal obtained by scanning the plurality of times. The information extraction means for extracting information using the maximum posterior probability estimation performs matrix calculation on the signal read by the plurality of scans using weighting at the transition between bright and dark portions of the image, and the weighting The optical information reading apparatus according to claim 1, further comprising a calculation unit that sets a value of the value in advance to reduce a calculation amount. Optical information reading method for reading a binary image having an arbitrary width in a direction orthogonal to the information arrangement direction while forming information by the length of the bright part and / or dark part of the image Because
Scanning the range of the arbitrary width a plurality of times out of phase with each other;
Retrieving information using a maximum posterior probability estimate for the signal read in the plurality of scans;
An optical information reading method comprising: Discriminating an arbitrary scanning line and a scanning line that is the same by shifting by one pixel of the optical information with respect to the signal read by the plurality of scans;
Taking out the scanning lines equally divided between the determined scanning lines;
Retrieving information using a maximum posterior probability estimate for the retrieved signal;
The optical information reading method according to claim 5, further comprising: In the maximum posterior probability estimation, matrix calculation is performed on the signal read by the plurality of scans using weighting at the transition between bright and dark portions of the image, and the weighting value is set in advance to calculate the amount of calculation. The optical information reading method according to claim 5, wherein the optical information reading method is reduced.