JP2002358529A - Method and device for processing image - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing method and its device by which a character and sign information on an object are read by a high resolution conversion processing based on a plurality of images which are extracted from a video obtained by photographing the mobile object. SOLUTION: Image areas p<0> -p<3> where the character or sign information concerning the object is included are respectively researched and taken-out based on a plurality of time sequential images from the video obtained by photographing the mobile object. The motion state for moving the object is assumed to be a uniform motion L, for example, and, then, positioning is performed to correct the positions of the image areas p<0> -p<3> to be at equal interval in times t0-t3 along the motion L. The positioned image areas p<0> -p<3> are superimposed to perform the high resolution processing so that the character and the sign included in a high resolution conversion image are read. Precision is enhanced in the high resolution processing through the use of the plurality of image areas which are positioned concerning the same object.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and an image processing apparatus for automatically reading characters and symbols in a video, and more particularly, to a method for extracting characters and symbols at high resolution from a plurality of extracted images. It relates to readable image processing.

[0002]

2. Description of the Related Art Conventionally, for example, in a school or a company, a camera is installed at an entrance of a building, and characters on a name tag attached to a body of a person who enters and exits are read. A system has been set up to manage the entry and exit of people into and out of the building based on the read letters of the name tag.

[0003] Further, in a production site such as a product production factory, when a shipment is made to flow on a belt conveyor and sorted,
Cameras installed at the site take pictures of the luggage flowing and moving on the belt conveyor and read the characters on the stickers affixed to the luggage from the video to automatically identify each luggage that moves. A distribution monitoring system that sorts by destination is used.

[0004] A technique for reading characters or symbols moving in an image is, for example, a logistics monitor provided with a belt conveyor 2, a lighting device 3, a camera 4, a recognition device 5, and a monitor as shown in FIG. This will be described using a system. In a production factory or the like, an image of the moving shipment 1 is photographed by the camera 4 installed beside the belt conveyor 2, and the characters of the label 7 attached to the shipment 1 are recognized by the recognition device 5 from the image. The logistics of the shipment 1 is monitored, such as reading and sorting the shipment 1 to a destination or the like.

A method for reading characters and symbols on the label 7 in a video is, for example, an image processing method called template matching (edited by the Institute of Television Engineers of Japan, "Image Engineering-Electronics of Images-", pp. 132-133). I was With this technology, the part where the characters in the object are written is matched with the character template that we have in advance as a "dictionary", the one that is most similar to the dictionary template is read, and the result is output I have.

In the reading method according to this technique, since the camera 4 and the shipment 4 are separated from each other, when the characters written on the label 7 become small, the characters themselves are crushed and become similar to other characters. This makes it difficult to recognize characters when reading. As a result, it is inevitable to target characters that are somewhat large. Therefore, when reading a character or a symbol on the label 7 by this technique, it is necessary to take a large image of the character or the symbol, and the shooting range of the camera 4 is limited. However, for example, in the physical distribution monitoring system, there are various sizes of the shipment 1 moving on the belt conveyor 2, and the position of the applied label 7 differs depending on the type of the shipment 1. For such reasons, it is necessary to increase the photographing range of the camera 4 as much as possible in order to detect the label 7 reliably.

To monitor a wider range using this technique, for example, Japanese Patent Application Laid-Open No. H11-284981
As disclosed in Japanese Unexamined Patent Application Publication No. H11-264, it is disclosed that a wide range is divided into a plurality of areas, and a plurality of cameras are used to allocate each area to each camera. However, in this method, not only the number of cameras increases, but also peripheral devices such as lighting devices and processing devices increase along with the increase.

[0008]

As described above, in the reading method according to the prior art, there is a trade-off between "the number of cameras" and "the readable range of characters and symbols". Therefore, when shooting a wide range with a small number of cameras, the resolution of the target character or symbol is low, and the information of the character or pattern is insufficient, so that the part is crushed or missing and it is not possible to read them There is a problem.

In order to solve this problem, a high resolution technology for increasing the resolution of an image by image processing has been developed (for example, Aoki et al., "Super Resolution Processing from Multiple Digital Images", 2nd Image Sensing Symposium, pp.65-70)
Have been. This high-resolution technique is to generate a high-resolution and clear image by using a plurality of images in a video and adding pixel information of each image.

Here, the principle of the high resolution processing in the high resolution technology will be described below. Higher resolution refers to modeling a process of forming an image and estimating a high-definition image that causes the formed image by solving the inverse problem. Specifically, for example, a high-resolution image is generated by the following solution. F (X, Y) is an ideal high-resolution image, and k (F, X, Y) is obtained by performing coordinate transformation.
The high-resolution image in the second frame and F _k (X, Y). That is, if the coordinate transformation formula is X = X _k (x, y), Y = Y _k (x, y) (1), the high-resolution image in the k-th frame is F _k (X, Y) = F (X _k (x, y), Y _k (x, y)) (2)

The k-th low-resolution observation image is _represented by G _k (i, j)
After all,

[0012]

(Equation 1)

Here, w (i, j: x, y) is a window function (PSF) determined by the spatial position and aperture characteristics of each CCD pixel. W (i, j; x, y) = 1, i−0.5 <x <i + 0.5, j−0.5 <y <j + 0.5 = 0 with respect to the integer value (i, j), etc. ). This means that each light receiving element is square and covers the image surface without any gap.
Note that the following discussion holds even if w (i, j; x, y) is a function of another form. From equations (2) and (3)

[0014]

(Equation 2)

Here, x _k (X, Y) and y _k (X, Y) are inverse transformations of the coordinate transformation of the equation (1). Also, (δ (x, y) / δ (X, Y))
Is the Jacobian of the variable transformation (a scale factor indicating how many times the area is increased for each location), and w _k (i, j; X, Y) = w (i, j: x _k (X, Y ), Y _k (X, Y)) (δ (x, y) / δ (X, Y)) (6)

[0016]

(Equation 3)

Is obtained. This equation shows that the image G _k (i, j) observed as the k-th frame and the high-resolution ideal image F
Represents the relationship (X, Y). Observed image G _k (i, j) when a given ideal image F (X, Y) of high resolution to be the group which is a problem to be solved because you seek. Equation (7) itself holds even if F (X, Y) is arbitrarily high resolution (X and Y are arbitrary real numbers), but if the resolution is too high, the number of unknowns is too large and the solution is unique. Not determined. Therefore, using a window function w (I, J; X, Y) defined earlier from a discrete image H (I, J) in which F (X, Y) has a value only on an integer lattice.

[0018]

(Equation 4)

It is assumed that Then F (X,
Y) is a step function, and the number of unknowns is reduced to the number of integer lattice points. Note that the discrete representation of the image may be another method, but this argument holds for any window function.

[0020]

(Equation 5)

Here, a high-resolution image H (I, J) that optimally satisfies Expression (11) in the sense of least squares may be obtained. That is,
Next evaluation function

[0022]

(Equation 6)

H (I, J) may be obtained so as to minimize the following. Through the processing described above, a high-resolution image is generated. However, with such high resolution technology,
Since a high-resolution image is generated by associating each pixel of each image with a pixel value and superimposing the pixel value, it is necessary to adjust the position of each image with an accuracy of one pixel or less. If the position is shifted, a clear high-resolution image cannot be obtained. For this reason, the alignment of each image is important. However, when the target image itself has a low resolution, it is difficult to perform strict alignment because the size of the target in the image is small.

Therefore, the present invention restricts the position of each target object to be read out in each image in accordance with the movement of the target object moving in the video, thereby achieving precise alignment throughout the entire image. It is an object of the present invention to provide an image processing apparatus and an image processing method that enable the image processing to be performed with higher resolution.

[0025]

According to the present invention, there is provided an image processing method for automatically reading target information relating to a moving target object from a video image of the moving target object. For each of a plurality of time-series images, a search step of searching for an image area in a predetermined range including the target information, and correcting a position of each of the searched image areas to a position suitable for movement of the target object. Thus, a positioning step of positioning each image area and a high-resolution image processing step of superimposing the aligned image areas and performing high-resolution processing are included.

The object information includes a character or a symbol, and in the searching step, in each of the images,
The image area is searched based on the layout information related to the character or the symbol. Further, in the positioning step, assuming that the movement of the target object is a constant-velocity linear motion, the position of each image region is corrected to a position on the constant-velocity straight line to perform positioning, and the position of the image region is adjusted. Is shifted from the movement position of the target object by more than a predetermined value, the image area is excluded from the target of the alignment processing.

According to the present invention, there is provided an image processing apparatus for automatically reading target information relating to a moving target object from a video image of a moving target object, wherein the target information is stored for each of a plurality of time-series images in the video. Search means for searching for an image area in a predetermined range including the image area, and position adjusting means for correcting the position of each of the searched image areas to a position suitable for the movement of the target object, and aligning the image areas. And a high-resolution image processing means for performing the high-resolution processing by superimposing the aligned image regions.

[0028]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment in which the present invention is applied to a character recognition processing device using a high resolution technology will be described below. [First Embodiment] As shown in FIG. 1, when a fixed camera 4 photographs a moving luggage 1 placed on a belt conveyor 2, an image of the luggage 1 displayed on a monitor 6 is: Suppose that it moves on the screen according to the movement of the belt conveyor. Here, a character-like area is cut out from a plurality of images including an image relating to the moving luggage 1 as a target, for example, based on arrangement information of a character string in the target. Since the position of the cut-out character area includes some errors due to the vibration of the luggage 1 and the like, errors between the respective images are accumulated and a large shift occurs as a whole.

Therefore, in the present embodiment, the movement of the object moving in the video is represented by some kind of movement in a very short time. For example, the movement of the load 1 on the belt conveyor 2 is expressed as "e.g. The position extracted from each image related to the label 7 affixed to the baggage 1 to be read is referred to as “constant linear motion”.
By constraining the position of the object so as to perform, precise alignment can be performed throughout the image. By such a positioning process, it is possible to perform a favorable high resolution process and to increase the resolution of the characters and symbols in the target, so that the characters and symbols written on the label 7 can be easily read.

FIG. 2 shows a block configuration of a character recognition processing device including a positioning process according to the first embodiment. An analog video input signal obtained from an image input device such as the camera 4 or a digital video input signal such as an output from an image recording device is supplied to the character recognition processing device 10.
When the input signal is an analog video signal, after the analog video signal is converted into a digital signal by the A / D converter 11, or when the input signal is a digital video signal such as an output from an image recording apparatus. Has an image control unit 12 which receives the digital signal as it is. Then, a first image storing the image input by the image control unit 12 is performed.
An image storage unit 13 and a second image storage unit 14 that stores results processed by the target search unit 15 and the alignment processing unit 16
Based on a given layout, a target search unit 15 for searching for a moving target, a positioning processing unit 16 for positioning a target in an image at each time, and a target processing unit 16 for positioning a target at each time A high-resolution processing unit 17 for performing a high-resolution process based on the result, and a character recognition unit 18 for reading a character string or a symbol from the high-resolution image are provided.

Hereinafter, the operation of the character recognition processing device 10 according to the first embodiment will be described in detail for each block. A / D conversion of input image An analog image obtained from an image input device such as the camera 4 is digitized by the A / D converter 11 and output to the image controller 12 at the subsequent stage. However, when the input video is a digital video, the configuration does not include the A / D converter 11. Control of A / D-converted image The image control unit 12 controls a digital video signal based on an output image from the A / D conversion unit 11 or an output image from an image recording device or the like. The image is stored and sent to the target search unit 15 at the subsequent stage. Image Storage The first image storage unit 13 stores image data relating to the digital video signal input to the image control unit 12. In addition, since the memory has a memory capacity enough to store image data for a certain period of time, it is possible to store the input image data even while the target search unit 15 in the subsequent stage performs the processing. -Storage of processing results The second image storage unit 14 stores results processed by the target search unit 15 and the positioning processing unit 16. Although the second image storage unit 14 is shown separately from the first image storage unit 13, the configuration as the storage unit may be one, and is shown as an individual configuration for convenience of explanation. Searching for Recognition Target The target search unit 15 searches whether or not an image based on the video signal sent from the image control unit 12 includes a target image.

First, it is checked whether or not an object is present in an image input from the image controller 12 and taken at a certain time t. Specifically, by performing binarization and labeling processing on the input image, FIG.
Detect character-like parts like In the figure, a character-like portion is shown in a round shape or an elliptical shape. Next, for example, layout information indicating an arrangement relationship of a target character string or the like is given in advance like a name tag as shown in FIG. 3B. At this time, the layout information does not need to be a character or a symbol, but is of such a degree that the positional relationship between them can be determined.

An area within a predetermined range in which the arrangement of the character-like portion is similar to the layout information is found. If the degree of similarity with the layout information is equal to or less than a certain threshold, a signal indicating “no target” is sent to the image control unit 12 and the image control unit 1
2 outputs the image at the next time stored in the first image storage unit 13 to the target search unit 15, and then outputs the target search unit 15
Performs a search on that image.

If the similarity of the image with the layout information is equal to or greater than a certain threshold, it is determined that the images are similar,
As shown in FIG. 3A, the coordinates (x1, y1), (x2, y) of the four corners of the area P having the predetermined range at the next time are shown.
2), (x3, y3), (x4, y4) and the image data relating to the area P are output to the second image storage unit 14 and stored. At the same time, a signal indicating “there is a target” is sent to the image control unit 12.

In this way, the image control unit 12
If the time t stored in the image storage unit 13 is from the time t0 to the time tn-1, the image at each time is retrieved from the target search unit 15
Are input sequentially. Then, the target search unit 15 performs a sequential search process on the sequentially input images. Here, the number n of the input images is the number used in the high-resolution processing unit 17.

Next, the position of the object from time t1 to tn-1 is tracked. Specifically, for example, the coordinates (x10, y10), (x20, y20), (x3
Object tracking is realized by template matching using the region p0 detected from (0, y30) and (x40, y40) as a template. The coordinates (x1i, y1i), (x2i, y2i), (x3i) of the four corners of the target area pi at each time ti obtained by tracking.
i, y3i), (x4i, y4i) and image data relating to the area pi are output to the second image storage unit 14 and stored. here,
If the number of processed sheets is n, i = 0 to (n-1).

FIG. 4 shows a specific example of the target area obtained by the search processing when n = 4. The target area is indicated by a square frame of p0 to p3. By repeating the above processing for the number of processed sheets n, the coordinates of the four corners of the target area at each time are obtained. On the basis of the coordinates of the four corners, the positioning processing unit 16 at the subsequent stage performs positioning on the target area of each image. When the processing in the target search unit 15 is completed, the image at each time stored in the second image storage unit 14 and the coordinates of the four corners of the target area in each image are stored in the alignment processing unit 16. Are sequentially input. Alignment of the object in the n images from the input time t0 to the time tn-1 is performed.

First, one of the n images is taken out, and
An image enlarged to a certain size (hereinafter referred to as a reference image) is created. Projection transformation for adjusting the positions of the remaining (n-1) images is determined so as to correspond to the images. The size of the enlargement is the size of the image generated by the high resolution processing. In general, when a correspondence of four points is given between two images, a projective transformation for linking the two images is determined. Thus, the four vertices in the m-th (m <n) input image are represented by (xm1, ym1), (xm2, ym2), (xm3, ym3), (xm
4, ym4), and the coordinates in the reference image are (xM1, yM1),
(xM2, yM2), (xM3, yM3), (xM4, yM4), and the projection transformation for transforming the reference image into the m-th input image is: xmi = (a1.xMi + a2.yMi + a3) / (a7.xMi + a8.yMi + a9 ) ymi = (a4.xMi + a5.yMi + a6) / (a7.xMi + a8.yMi + a9) (Since there are four vertices, i = 1, 2, 3, 4) (13)

The coordinates of the four corners of each image obtained by the object search unit 15 are substituted into equation (13). This gives eight linear equations to the nine unknowns of the conversion coefficients a1 to a9. Equation (13) shows that the scale invariance that does not change even when a1 to a9 is multiplied by a constant is calculated. So, for example, a9
Given the constraint that = 1, the solution is uniquely determined. Through the above processing, projective transformation for transforming the reference image into the m-th input image is obtained. If this inverse transformation is performed on the m-th input image, the position matches the reference image.

Here, actually, since the resolution of the target area of each image is low, the obtained position of the target area includes an error. Therefore, the alignment is corrected according to the movement of the target object. When the moving target object is viewed at a minute time interval, it can be seen that the moving target object is performing a predetermined predetermined movement. By correcting the position of the target area detected at each time so as to satisfy the movement, the position of the target area can be strictly adjusted.

Specifically, for example, as shown in FIG. 4, when the positions of the target areas p0 to p3 by the respective inverse projection transformations at times t0 to t3 are represented, as shown in FIG. The motion of the target area in time is applied to a predetermined motion.
For example, the motion of the target object is regarded as a uniform linear motion. FIG.
Is assumed to be moving like the arrow indicated by the broken line L. The figure shows the motion of the target object as viewed on a screen on the monitor 6.

First, among the points corresponding to the four corner positions of the target area at each time, the upper left point of the areas p0 to p3 is assumed to be the position representative of the area, and their coordinates are indicated by circles in FIG. (X10, y10), (x11, y11),
(x12, y12) and (x13, y13). Therefore, as shown in FIG. 6, the representative position of each target area as indicated by a circle in FIG. 5 is corrected so as to be aligned on a straight line L representing the motion of the target object. As shown, each coordinate is (xm10, y
m10), (xm11, ym11), (xm12, ym12), (xm13, ym1
Modified in 3). However, here, each coordinate is merely located on the straight line L, and has not yet been corrected to the constant velocity motion.

Next, in order to assume that each coordinate is moving at a constant speed indicated by a straight line L as shown in FIG. 7, the coordinates are corrected so that the movement amounts at each time interval become equal. Using projective transformation according to equation (13), recalculation is performed on the image data of each target area. The coordinates after the correction are (xM10, yM10), (xM11, yM11), (xM12, yM1) as shown in FIG.
2), (xM13, yM13), and the coordinates of each area are corrected such that the target areas p0 to p3 have a uniform linear motion along the straight line L.

As described above, by constraining the coordinates at each time by the entire motion, errors occurring between images at each time are not accumulated, so that strict positioning without a large shift can be performed. Also, the case where the projective transformation is obtained only by the correspondence of the four corner points without being restricted by the movement of the object is included. As described above, the position of the target at each of the times t0 to t3 is adjusted to form a superimposed image, and the image is output to the high resolution processing unit 17. High-resolution processing of superimposed images The high-resolution processing unit 17 performs high-resolution processing on the n images generated by the alignment processing unit 16. The resolution enhancement performed here is a model of the image forming process,
It is to estimate a high-definition image that causes the image formed by solving the inverse problem, and specifically,
This is a high resolution process based on equations (1) to (12). H (I, J) that minimizes the evaluation function represented by equation (12) is obtained. The high-resolution image generated by such processing is output to the character recognition unit 18. Character Recognition from High Resolution Image The character recognition processing unit 18 performs character recognition using the image output from the high resolution processing unit 17.

Specifically, for example, a dictionary is prepared in advance in order to recognize patterns having different sizes and phases of characters (kanji, numerals, alphabets, etc.) which may be included in the character layout. Therefore, template matching is performed between the characters in the image that has been subjected to the high resolution processing and the dictionary image, and those having a high degree of matching are output as recognition results of the characters and the like. As a method of template matching, the equation (14) is calculated for the i-th pixel X _i , Y _i in the input image (X, Y), and the smaller the degree of matching M, the more similar the area. However, M ≧ 0.

[0046]

(Equation 7)

The recognition result of characters and the like by the above processing is
The output is output to a separately prepared sorting device downstream of the belt conveyor 2. The sorting device automatically sorts the luggage 1 as a target object,
The distribution monitoring system can be constructed, for example, in such a way that the worker sorts out the luggage 1 by outputting the data to the user. [Second Embodiment] FIG. 8 shows a block configuration of a character recognition processing device 10 according to a second embodiment.

In the character recognition processing device 10 according to the first embodiment shown in FIG. 2, an image selection processing unit 19 is added at the subsequent stage of the positioning processing unit 16. In the first embodiment, high resolution processing is performed using n input images, which are images at each time from the time t0 to tn-1 when the target is found. However, there is a high possibility that the input n images include images that are not suitable for high-resolution processing, such as images that have not been properly aligned. When such an image is used for the high resolution processing, it may have a bad influence.

Therefore, in the second embodiment, good resolution can be obtained by removing an image having a large displacement from n images. Specifically, for example, the template matching shown in Expression (12) in the above-described high-resolution processing is performed between each of the aligned images and the reference image. It is considered that the displacement is large. Therefore, after the matching degree is calculated by the image selection processing unit 19 and the image whose value is larger than the threshold value is removed, the high resolution processing is performed based on the image with relatively small positional deviation.

Therefore, in the high resolution processing, an image unsuitable for the processing is removed from the processing target, so that the processing can be performed with higher accuracy. Until now, the explanation has been given using the physical distribution monitoring system as an example, but the present invention is not limited to this, and it is necessary to recognize target information attached to moving objects, such as a system for checking in and out of people in buildings. In addition, the high-resolution image processing using a plurality of images according to the present embodiment can be applied.

The object information is not limited to characters or symbols, but may be a figure in a specific shape in an object. In this case, a graphic recognizing unit is provided instead of the character recognizing unit at the subsequent stage of the high resolution processing unit. As an operation of the character recognition processing apparatus to which the resolution increasing process according to the present embodiment is applied, an example in which an object to be searched is moving according to a constant velocity linear motion has been described. If it is known that the user moves along a predetermined route within a certain range, each target region may be modified so that the route is restricted as the search target motion.

In the above, an example in which one target object is searched for in a screen shot by a camera has been described. However, two or more target objects can be searched at the same time. By preliminarily determining the motion of each target object, it is possible to monitor a wide range of each target object based on an image captured by one camera.

Further, when the screen shot by the camera fluctuates, even if the target object is stationary,
By constraining each target area to a stationary motion as a fixed path, a clearer high-resolution image can be created. (Supplementary Note 1) An image processing method for automatically reading target information relating to a moving target object from a video image of the moving target object, the method comprising:
A search step of searching for a predetermined range of image areas including the target information, and correcting the position of each of the searched image areas to a position suitable for the movement of the target object; An image processing method comprising: performing a positioning step; and performing a high-resolution image processing step of performing a high-resolution process by superimposing the aligned image areas. (Supplementary note 2) The supplementary note 1, wherein the target information includes a character or a symbol, and in the searching step, the image area is searched in each of the images based on layout information on the character or the symbol. The image processing method described in the above. (Supplementary Note 3) In the positioning step, it is assumed that the target object is moving every minute time, and the position of the image area is corrected according to the moving position of the target object corresponding to the time to perform positioning. 3. The image processing method according to claim 1 or 2, wherein (Supplementary Note 4) In the positioning step, the position of each of the image regions is corrected to a position on the constant velocity straight line, and the position is adjusted, assuming that the movement of the target object is a constant velocity linear movement. 4. The image processing method according to claim 3, wherein (Supplementary Note 5) In the positioning step, when the position of the image area is deviated from a movement position of the target object by more than a predetermined value, the image area is excluded from a target of the positioning processing. 5. The image processing method according to any one of 1 to 4. (Supplementary note 6) The image processing method according to any one of Supplementary notes 2 to 5, further comprising a character recognition processing step of recognizing the character or symbol information based on the high-resolution image. (Supplementary Note 7) An image processing apparatus that automatically reads target information related to a moving target object from a video image of the moving target object, the image processing device including:
Searching means for searching for a predetermined range of image areas including the target information; correcting the position of each of the searched image areas to a position suitable for the movement of the target object; An image processing apparatus, comprising: a positioning unit that performs a positioning process; and a high-resolution image processing unit that performs a high-resolution process by superimposing the aligned image regions. (Supplementary Note 8) The supplementary note 7, wherein the target information includes a character or a symbol, and the search unit searches the image area in each of the images based on layout information on the character or the symbol. The image processing apparatus according to any one of the preceding claims. (Supplementary Note 9) The positioning means corrects the position of the image area according to the movement position of the target object corresponding to the time, assuming that the target object moves every minute time. An image processing apparatus according to claim 7 or 8, wherein (Supplementary Note 10) The positioning means performs the positioning by correcting the position of each of the image regions to a position on the constant velocity straight line, assuming that the movement of the target object is a constant velocity linear movement. 10. The image processing apparatus according to claim 9, wherein (Supplementary Note 11) The positioning unit includes an image selection unit that excludes the image region from a target of the positioning process when the position of the image region is shifted from a moving position of the target object by more than a predetermined value. The image processing apparatus according to any one of supplementary notes 7 to 10, wherein: (Supplementary Note 12) The image processing apparatus according to any one of Supplementary Notes 8 to 11, further comprising a character recognition processing unit configured to recognize the character or symbol information based on the high-resolution image.

[0054]

The effect of the present invention is that the position of a moving object including characters and symbols in a video at each time can be strictly adjusted by constraining the movement of the object. The conversion process can be performed well, and characters and symbols can be read automatically. Thereby, the photographing range of the camera can be widened. For example, when the present invention is applied to a distribution monitoring system, a plurality of belt conveyors can be monitored by one camera, leading to a significant cost reduction of the system.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of a physical distribution monitoring system.

FIG. 2 is a diagram showing a block configuration according to a character recognition processing device to which the first embodiment is applied.

FIG. 3 is a diagram illustrating a search for a recognition target based on layout information.

FIG. 4 is a diagram showing a position of a recognition target at each time displayed on a monitor.

FIG. 5 is a diagram illustrating a state in which a representative point of the target image for correcting the motion of the recognition target in a very short time is specified.

FIG. 6 is a diagram for explaining correction of an image position by a motion of a recognition target by arranging representative points of each target image on a straight line.

FIG. 7 is a diagram showing a state in which each target image has been corrected so that the amount of movement at each time is equal so that the recognition target has a uniform linear motion.

FIG. 8 is a diagram illustrating a block configuration according to a character recognition processing device to which the second embodiment is applied.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Luggage 2 ... Belt conveyor 3 ... Lighting device 4 ... Camera 5 ... Recognition device 6 ... Monitor 7 ... Label 10 ... Character recognition processing device 11 ... A / D converter 12 ... Image control unit 13 ... First image storage unit 14 ... Second image storage unit 15 ... Object search unit 16 ... Positioning processing unit 17 ... High resolution processing unit 18 ... Character recognition unit 19 ... Image selection processing unit

Continued on the front page (51) Int.Cl. ⁷ Identification FI FI Theme Court II (Reference) // H04N 5/915 H04N 5/91 K (72) Inventor Morito Shiohara 4-1-1 Kamiodanaka, Nakahara-ku, Kawasaki City, Kanagawa Prefecture No. 1 F-term in Fujitsu Limited (reference) 5B057 CA12 CA16 CB12 CB16 CE03 5C052 AA00 DD04 5C053 FA11 GB19 HA29 KA01 KA24 LA01 5C054 AA01 EJ07 FC13 HA03 5L096 FA69 HA04 JA03

Claims (5)

[Claims] 1. An image processing method for automatically reading target information relating to a moving target object from a video image of a moving target object, wherein the predetermined information includes the target information for each of a plurality of time-series images in the video. A search step of searching for a range of image areas; a position adjusting step of correcting the position of each of the searched image areas to a position suitable for movement of the target object, and aligning each image area; A high-resolution image processing step of performing high-resolution processing by superimposing the aligned image areas. 2. The method according to claim 1, wherein the target information includes a character or a symbol, and in the searching step, the image area is searched in each of the images based on layout information on the character or the symbol. 2. The image processing method according to 1. 3. In the positioning step, the position of each of the image areas is corrected to a position on the constant velocity straight line, and the position is adjusted, assuming that the movement of the target object is a constant velocity linear movement. The image processing method according to claim 1. 4. In the positioning step, when the position of the image area deviates from a moving position of the target object by more than a predetermined value, the image area is excluded from a target of the positioning processing. The image processing method according to claim 1. 5. An image processing apparatus for automatically reading target information related to a moving target object from a video image of the moving target object, the predetermined information including the target information for each of a plurality of time-series images in the video. Searching means for searching for an image area in a range, correcting the position of each of the searched image areas to a position suitable for movement of the target object, and aligning each image area; An image processing apparatus comprising: a high-resolution image processing unit that performs high-resolution processing by superimposing the aligned image regions.