JP3403697B2 - Image processing method and apparatus - Google Patents

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 apparatus, and more particularly, to a relative moving speed between an observation point and a target object and a target from time-series image data taken by using a line sensor camera or the like. The present invention relates to a method for measuring the length of an object, a pattern recognition method using the method, and devices thereof.

[0002]

2. Description of the Related Art As a method of measuring the velocity of a moving object using a plurality of sensors, there is a method using two photocells. In this method, two photocells are installed in parallel at a distance from each other, and the speed is obtained from the time difference when a moving object crosses in front of the phototube, and it is widely used for speed measurement of automobiles.

On the other hand, there are the following methods for measuring the speed and length of a moving object from an image. "Traffic flow measurement using double slit image" (reference (1), IEEJ Road Traffic Research Society material RTA94-5, 1994) and "traffic measurement using double slit camera" (reference (2), image) The Institute of Electronics Engineers, 26-3, 1997). With these methods,
A spatio-temporal image is used in which a slit is artificially provided in an image captured by a general video camera and the images on the slit are connected. The passage speed and the length of the object are measured by providing two slits and determining the time for the object to pass between the slits.

[0004]

However, the above-mentioned prior art has the following problems. Considering the problem that the observation side or target object moves relatively on a predetermined orbit and relative speed measurement, target object length measurement, and target object pattern recognition processing are considered, the conventional photocell is used. In the conventional measurement, the measuring device must be installed so as to sandwich the target object. Moreover, since an image cannot be acquired by this method, it is unknown what target object has passed in front of the sensor. Therefore, the target object must be visually confirmed by hand.

On the other hand, in the conventional video image, since the two slits are not parallel to each other, it is necessary to fix the depth in advance and separately calculate the actual moving distance between the slits. In addition, a general video camera can shoot only 30 frames / second, and it is impossible to obtain speed measurement accuracy when the target object moves at high speed. Further, there is no method for measuring the relative moving speed with respect to the target object from the image by using the camera installed on the moving object.

Therefore, one of the objects of the present invention is to solve the problems of the prior art and to propose a new method for measuring velocity and length, a pattern recognition method, and their devices. , Measurement of relative moving speed when the observation point and the target object move relative to each other, measurement of moving speed of the target object and length of the target object when the observation point is fixed, and of the target object in those cases It is to provide a technique capable of easily performing pattern recognition processing.

[0007]

Means for solving the problems Means for solving the above problems are as follows.

[0008]

[0009]

[0010]

[0011]

According to a first aspect of the present invention, when a target object having a specific shape pattern recognizable as an image moves on a predetermined orbit relative to an observation point, the target object is recognized. An image processing method where the line axis is
It traverses the trajectory of a target object that is parallel and moves relatively.
Was installed at a predetermined angle with respect to the relative movement direction
Synchronize multiple line image acquisition devices to capture a constant line
Time series image data taken in time series with shadow cycle time
And by calculating the similarity,
Correlate the image of the target object between the time-series image data
And the shift amount of the target object between the associated images and
The line capture cycle time and the pair between each line image acquisition device
Determining the moving time of the elephant object, and the moving time
And the image acquisition of each line image acquisition device on the orbit
The distance between the obtained position, determining a speed of relative movement between the observation point and the target object from the predetermined angle, the time-based
Dividing the time axis of the string image data in the relative movement speed, by correcting the scale of the time axis direction, a step to match the scale of the time-series image data prepared in advance templates of the specific shape pattern , Recognizing the specific shape pattern in the time-series image data by calculating a similarity between the time-series image data having a matched scale and the template of the specific shape pattern. The gist of the image processing method is as follows.

Further, the invention according to claim 2 is an image processing method for recognizing a target object having a specific pattern recognizable as an image when the target object moves on a predetermined trajectory. , The line axes become parallel and move
To the direction of movement so that it traverses the trajectory of the moving target object
A plurality of line image acquisition devices installed at a predetermined angle
Synchronized with each other and take time-series images with a constant line shooting cycle time.
The step of acquiring each shadowed time series image data, and
By calculating the similarity between each time series image data
, The object between the time series image data on the time axis
Correspond to the body, or with each of the time series image data
An image change whose difference from the previously acquired background image is more than a threshold value.
Of the image in each of the time-series image data.
Based on the image change point, the time-series image data
To associate the target object with, or each of the time series
Each line image in image data and each one cycle before
Detects image change points where the difference from the line image exceeds a threshold
The image change point in each of the time-series image data
Based on the time axis, the target between the time series image data
By associating the objects, each line of the target object
Calculate the amount of deviation between the image acquisition positions of the image acquisition device, and
The moving time is calculated from the amount of deviation and the line shooting cycle time.
Step, the travel time, and the orbit
The distance between the image acquisition positions of each line image acquisition device,
A step of obtaining the moving speed of the target object from a predetermined angle, and dividing the time axis of the time series image data by the moving speed .
By correcting the scale in the time axis direction,
Calculating the similarity between the time-series image data in advance and the step of providing been matched the scale of the template of the specific shape pattern, template matching the specific shape pattern and the time-series image data of the scale Accordingly, the gist of the image processing method is characterized by including the step of recognizing the specific shape pattern in the time-series image data.

[0014]

[0015]

[0016]

[0017]

[0018]

[0019]

[0020]

Further, according to the invention of claim 3 , when a target object including a specific shape pattern recognizable as an image moves on a predetermined orbit relative to an observation point, the target object is detected. An image processing device that recognizes lines
Trajectory of the target object with parallel axes and relative movement
Installed at a predetermined angle to the relative movement direction.
Of multiple line image acquisition devices
Shooting operations are performed in chronological order with the shooting cycle time, and each time series is
Image acquisition means to acquire image data and calculate similarity
By doing so, the target object is
Correlating images and target objects between the correlated images
Each line image from the deviation amount of the line and the line imaging cycle time
Time calculating means for obtaining the moving time of the target object between the acquisition devices
And the travel time and each line image on the orbit
The distance between the image acquisition positions of the acquisition device and the predetermined angle
Velocity to find relative moving velocity between target object and observation point from
Calculating means and the time series image acquired by the image acquiring means
Dividing the time axis of the image data in the relative movement speed determined by the speed calculating means, by correcting the scale of the time axis direction, before said image acquisition means acquires
And scale correcting means for matching the scale of the template of the serial time-series image data and prepared in advance the specific shape pattern, the similarity between the scale matched the time-series image data and the template of the specific shape pattern The gist of the image processing apparatus is to include a pattern detection unit that detects and recognizes the specific shape pattern by calculating

Further, in the invention according to claim 4, when the target object including a specific shape pattern recognizable as an image moves on a predetermined trajectory, the target object is recognized at a stationary observation point. The device has a flat line axis
Move in a row and traverse the trajectory of the moving target object.
Multiple lines installed at an angle to the moving direction
During a constant line shooting cycle while synchronizing the image acquisition device
Time-series image capturing operation to acquire each time-series image data.
Image acquisition means to be obtained and the kind between the time series image data
By calculating the similarity, each time series image on the time axis
Correlate the target object between image data, or
Difference between each time series image data and the background image acquired in advance
Image change points where
Based on the image change point in the image data,
Correspondence of the target object between each time series image data
Alternatively, each of the time series image data
The difference between the in-image and the line image each one cycle before is less than the threshold value.
The upper image change point is detected, and each time series image data is detected.
On the time axis based on the image change point in
By associating the target object between the column image data,
The image acquisition position of each line image acquisition device of the target object
Calculate the amount of misalignment between units
A time calculating means for obtaining a moving time from the cycle time, and
Travel time and each line image acquisition device on the orbit
From the distance between the image acquisition positions of the
Speed calculation means for calculating the moving speed of the object, and the image acquisition
Means divided by the moving speed obtained by said speed calculating means time axis of the time-series image data acquired by correcting the scale of the time axis direction, the image
Templates and scale correction means, the scale matched to the time-series image data of the specific shape pattern to match the scale of the template of the specific shape pattern of the image acquisition means is prepared in advance and the acquired time-series image data An image processing apparatus is characterized in that the image processing apparatus is provided with a pattern detection unit that detects and recognizes the specific shape pattern by calculating the similarity between the pattern shape and the pattern shape.

[0023]

[0024]

[0025]

[0026]

[0027]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings.

<Embodiment 1> Embodiments of a velocity measurement and pattern recognition method according to the first embodiment of the present invention, and an apparatus thereof will be described. In the present embodiment, as an example of image processing using a plurality of line sensor cameras, a measurement system in which two line sensor cameras are installed is mounted on a moving vehicle, and moving speed measurement of the moving vehicle moving on a predetermined track, The stationary object is recognized as a stopped vehicle and the stopped vehicle is recognized. Here, the predetermined track is, for example, a lane on a road. Then, at this time, the moving vehicle and the stationary object move relative to each other, and the stationary object moving on a predetermined trajectory in the system with the moving vehicle as a reference is observed by the line sensor camera. The relative moving speed is measured.

FIG. 1 shows an apparatus for carrying out the velocity measurement and pattern recognition method according to the first embodiment of the present invention. The information processing device 101 shown in FIG.
A storage device 103, a display device 104 such as a display, and a measuring device 105 are connected via 02. However, in the storage device 103, a measurement value storage unit 106 and an image storage unit 107 are provided. A line sensor camera A 108, a line sensor camera B 109, and a synchronization device 110 are provided in the measuring device 105. Here, the synchronizer 110 is, for example, a pulse generator or the like. The time-series image data from the line sensor camera is imaged at regular line imaging cycle times by synchronizing the two cameras using the synchronization device 110.

The line sensor camera used in the present embodiment is a monochrome line sensor camera, each pixel has 8-bit 256 gradations, the number of pixels in each line is 2048, and the photographing time interval for each line, that is, the line photographing cycle time is 10.
It is 0 microsecond.

FIG. 2 is a diagram showing an example of a speed measuring method, a pattern recognizing method, and a measuring system in the apparatus according to the present embodiment. 201 is a stopped vehicle 1, 202 is a stopped vehicle 2, 203 is a moving vehicle, and 204 is a moving vehicle. Is the moving direction, 205 is the line sensor cameras A and 20 installed in the moving vehicle 203.
A line sensor camera B 6 is also installed on the moving vehicle 203.

FIG. 3 is a view of the measurement system of FIG. 2 seen from above. 301 is a moving vehicle, 302 is a moving direction, 303 is a line sensor camera A, 304 is a line sensor camera B,
305 is an angle θ, 306 is a line axis distance L, 307 is a moving distance Lr, 308 is a stopped vehicle 1, and 309 is a stopped vehicle 2. The moving direction of the moving vehicle 301 is the moving direction 3
02, and the line sensor camera A with respect to the moving direction 302.
303 and the line sensor camera B304 have an angle θ305
Suppose you are leaning at. At this time, the actual moving distance Lr307 when the stopped vehicle relatively moves between the line axes is calculated as Lr = L / sin (θ).

FIG. 4 is a flow chart of the speed measurement and pattern recognition method of this embodiment. Step 4 shown in FIG.
The processing from 01 to step 406 is processing executed by the information processing apparatus 101 such as the CPU of the personal computer.
The speed measurement and pattern recognition method of this embodiment will be described below with reference to the flowchart of FIG.

First, the moving speed of the relative moving vehicle is measured. First, the time series image data A411 taken by the line sensor camera A303 is divided into strip images (step 401). FIG. 5 shows an example of the time-series image data of the line sensor camera and the divided strip image.
Is the time-series image data A of the line sensor camera A, 502
Is time series image data B of line sensor camera B, 503
Is a strip image, 504 is a matching area, 505 is a strip width W, and 506 is a moving time Tv. The number of lines in the time axis direction of the strip image is preferably about 10 to 20, and the number of pixels in one line is 2,048. Therefore, in this case, the number of pixels in one strip image is about 20,480 to 40,960.

Next, the strip image and the line sensor camera B3
The correspondence with the time series image data B412 of No. 04 and the time shift amount are calculated (step 402). From FIG. 5, the line sensor camera B3 is arranged before the line sensor camera A303.
Since the stopped vehicle crosses in front of 04, a region 504 that coincides on the time-series image data B corresponding to the strip image 503 exists ahead on the time axis 507. Therefore, the similarity is calculated by using the difference calculation and the correlation coefficient calculation (see “Image Processing Handbook” (Reference (3), Shokoido, 1987)) only in the negative direction of the time axis, and Area 50 that was matched using the value
4 is detected, and the time shift amount is calculated.

Here, the strip image 503 and the time-series image data B502 divided based on the time-series image data A501.
A method of associating with will be described. For association, the cross-correlation value between both images is calculated, and the point having the highest correlation value within the predetermined range and having the highest correlation is detected. When the number of pixels on one line of a strip image is m and the number of lines is n, the brightness (gradation) of each point inside the strip image is Ia.
Let (i, j). In addition, the brightness of each point in the partial area for n lines starting from time t in the time series image data B412 is I
Let b (t, i, j). Assuming that the average and variance of brightness for each region are μa, μb (t), σa ² and σb (t) ² , the cross-correlation value c (t) between these regions is given by the following equation.

[0037]

[Equation 1]

This c (t) takes a value in the range of -1 to 1, and becomes closer to 1 as the degree of similarity between both partial areas is higher. Therefore, with t as a variable, that is, the strip image 503 is shifted in the time axis direction on the time series image data B502, and t at which the value of c (t) is closest to 1 is detected, and the time series image data B5 is detected.
The matched area 504 in 02 is determined.

Instead of calculating the cross-correlation value, the difference value d (t) according to the following equation may be used.

[0040]

[Equation 2]

The closer this d (t) is to 0, the higher the similarity between the two images.

If the number of lines in the strip image is too small, there is a high possibility that an error will occur in the correspondence between the images. Conversely, if the number of lines is too large, the calculation time for calculating the similarity becomes too long. Not only that, the change of the correlation value or the difference value becomes too gradual, and the accuracy of the association decreases. Therefore, when the number of pixels in one line is about 2048 as in the present embodiment, the number of lines in the strip image is set to 1
It is desirable to set it to about 0 to 20.

Next, the moving speed is calculated using the time shift amount of each strip image and the corresponding moving distance (step 403). Specifically, the time shift amount, that is, the moving time Tv is calculated as Tv = P × Tc by using the moving pixel number P of the strip image and the shooting time interval Tc of the line sensor camera. Therefore, the moving speed Vm of the moving object is calculated as Vm = Lr / Tv.

If the above speed measurement processing is performed on all of a plurality of strip images and the obtained plurality of measurement results are averaged or a representative value is obtained by majority decision,
The measurement accuracy can be improved. That is, each strip image included in the time-series image data A captured by the line sensor camera A is associated with the time-series image data B captured by the line sensor camera B to obtain the time shift amount, and the speed is calculated. calculate. Then, based on the speed measurement result values corresponding to the number of strip images thus obtained, an average value is calculated, or a representative value is obtained by a majority vote or the like.

As described above, the moving speed of the moving object can be measured using the two line sensor cameras.

Next, pattern recognition of the geometrical specific pattern of the moving object is performed from the time-series image data,
A process of recognizing a moving object will be described.

First, the time-series image data and the template of the specific pattern prepared in advance in a database or the like are extracted using the moving speed, and the scale in the time axis direction is corrected using the already calculated moving speed. Then, the degree of similarity is calculated (step 405). Here, the wheel is used as the specific pattern to be detected.

FIG. 6 shows an example of a template of a specific pattern, and 601 is a template of the specific pattern.
2 is the length Xt in the time axis direction, 603 is the length Yt in the line axis direction, 604 is the template of the corrected specific pattern, 605 is the length X in the time axis direction, and 606 is the length Y in the line axis direction. is there. Template 601 of specific pattern
Is the time-series image data captured at the time interval Ttc, X = Xt × Ttc / Tc, and the scales in the time axis direction can be matched. On the other hand, the scale in the line axis direction changes depending on the depth distance from the line sensor camera at the time of shooting. Therefore, the template of the specific pattern is enlarged / reduced in the line axis direction to match the scale. Then, the degree of similarity is obtained by the above-described difference calculation or cross-correlation value calculation (see reference (3)), and when a degree of similarity higher than a predetermined degree of similarity is obtained, it is determined that the patterns match. Here, the template of the specific pattern is corrected in order to match the scale, but the time-series image data or both may be corrected.

Next, the pattern is recognized using the similarity,
A stopped vehicle is recognized (step 406). Wheels that are a specific pattern are recognized by performing threshold processing, that is, processing based on the determination whether the similarity calculated in step 405 is greater than or equal to a predetermined threshold. Therefore, if the wheels are detected in the time-series image data, it means that the stationary object is a stopped vehicle. As described above, the stationary object can be recognized from the time-series image data using the two line sensor cameras.

In this embodiment, the observation point is moved and the target object is stationary, but even if both are moving, the relative moving speed can be measured by the same method. Further, although the example in which the wheels are used as the specific pattern has been described above, it is also possible to perform pattern correction in the time axis direction according to the moving speed and then recognize the pattern such as the license plates attached to the front and rear surfaces of the vehicle. It is possible.

<Embodiment 2> Embodiments of a velocity and length measurement and pattern recognition method according to a second embodiment of the present invention, and an apparatus thereof will be described below. In this embodiment, as an example of a plurality, two line sensor cameras are installed at fixed observation points, and velocity measurement, length measurement, and pattern recognition of a moving object moving on a predetermined trajectory are performed. Here, the predetermined trajectory is, for example, a lane when the moving object is a vehicle, or a sidewalk when the moving object is a pedestrian. The measuring device of this embodiment is the same as that of the first embodiment (see FIG. 1).

FIG. 7 is a diagram showing an example of a velocity / length measuring and pattern recognizing method and a measuring system in the apparatus according to the present embodiment. 701 is a moving vehicle, 702 is a line sensor camera A, and 703 is a line sensor camera. B and 704 are measuring devices.

FIG. 8 is a view of the measurement system of FIG. 7 seen from above. 801 is a moving vehicle, 802 is a moving direction, 803 is a line sensor camera A, 804 is a line sensor camera B,
805 is the angle θ, 806 is the line axis distance L, and 807 is the movement distance Lr. It is assumed that the moving direction of the moving vehicle 801 is a moving direction 802, and the line sensor camera A 803 and the line sensor camera B 804 are inclined at an angle θ805 with respect to the moving direction 802. At this time, the actual moving distance Lr807 when the moving vehicle passes between the line axes is calculated as Lr = L / sin (θ).

FIG. 9 is a flowchart of the speed measurement, length measurement, and pattern recognition method of this embodiment. The processing of steps 901 to 906 shown in FIG. 9 is executed by the information processing device 101 such as the CPU of a personal computer. Processing. The speed measurement, length measurement, and pattern recognition method of this embodiment will be described below with reference to the flowchart of FIG.

First, the moving speed of the moving vehicle is obtained. First, the line sensor camera A 803 and the line sensor camera B 804 extract an image pattern including the moving vehicle 801 from each of the time-series images 911 and 913 that are time-sequentially imaged at a constant line imaging cycle time by the line sensor camera A 803 and the line sensor camera B 804 (step 901). The images captured by the line sensor camera A 803 and the line sensor camera B 804 in advance when there is no moving vehicle are the background image A 912 and the background image B 91.
It is stored as 4 in the image storage unit 107.

FIG. 10 is a diagram showing an example of a time series image of the line sensor camera. 1001 is time series image data A of the line sensor camera A, 1002 is time series image data B of the line sensor camera B, and 1003 is A background image A for the line sensor camera A, 1004 is a background image B for the line sensor camera B, and 1005 is a time axis. The background image A1003 and the background image B1004 are images having a width of one line or several lines. In this embodiment, since the observation points are fixed, the time series image data A
As in 1001 and time-series image data B1002, an image in which the moving vehicle 801 exists on a continuous background image is obtained. Therefore, the difference calculation with the background image and the correlation coefficient calculation (see Reference (3)) are performed for each line or every several lines, and the threshold processing and the labeling processing (see Reference (3)) are performed based on the obtained values. Thus, an image pattern including a region different from the background image, that is, a moving object can be extracted. However, since the background image changes depending on the illumination condition, the background image may be updated at any time. In addition,
For the difference calculation and the correlation coefficient calculation, the cross-correlation value c
(t) or the definitional expression of the difference value d (t) can be used.

Next, the extracted image patterns are associated with each other in time series images to obtain the time shift amount of the image patterns (step 902).

FIG. 11 is a diagram showing an example of an image pattern extracted from each time series image data. 1101 is an image pattern A in the time series image data A and 1102 is an image pattern in the time series image data B. B, 1103 is a moving time Tv, 1104 is a shooting time Tm, and 1105 is a time axis. First, each image pattern is associated. Specifically, they are associated based on the difference calculation between image patterns and the correlation coefficient calculation (see Reference (3)). After associating, the time shift amount of the corresponding image pattern, that is, the moving distance Lr807
The moving time Tv1103 is calculated.

Next, the moving speed using the time shift amount of the image pattern and the corresponding moving distance is obtained (step 903). Specifically, the time shift amount, that is, the moving time Tv is
Using the number P of moving pixels of the image pattern in the time-series image data and the shooting interval Tc of the line sensor camera, Tv = P × Tc is obtained. Therefore, the moving speed Vm of the moving object is calculated as Vm = Lr / Tv. As described above, it is possible to measure the moving speed of the moving object using the two line sensor cameras.

Next, the length of the moving object is obtained using the shooting time and the moving time of each image pattern (step 90).
4). The image pattern shooting time Tm1104 is the time between the start point and the end point of the image pattern on the time axis. Therefore,
The length Lm of the moving object is calculated by Lm = Vm × Tm. Generally, when the angle θ 805 based on the moving direction 802 is other than 0, not only the side surface but also the front surface is photographed as shown in FIG. 10, so that an accurate object length in the moving direction cannot be obtained. However, as shown in FIG. 12, even when the angle θ is other than 0, when only the side surface is reflected in the time-series image data and the side surface is parallel to the moving direction, an accurate object length in the moving direction can be obtained. . Therefore, when measuring the length of a moving object, the angle θ80
A time-series image when 5 is set to 0 is used (see FIGS. 13 and 1).
4).

Next, pattern recognition of the geometrical specific pattern of the moving object is performed from the time-series image data,
It recognizes moving objects.

First, the time-series image data and the template of the above-described specific pattern prepared in advance in a database or the like are taken out by using the moving speed, corrected and scaled, and the similarity is calculated (step 905). Here, the wheels are used as the specific pattern to be detected as in the first embodiment. Similar to step 405 shown in the first embodiment,
In this embodiment as well, the degree of similarity can be obtained.

Next, the pattern is recognized by using the similarity,
A moving vehicle is recognized (step 906). Step 90
By performing threshold processing on the similarity of 5, the wheels that are the specific pattern are recognized. Therefore, the moving object shooting time 11
If a wheel is detected in the time series image data in 04, it means that the moving object is a moving vehicle. As described above, the moving object can be recognized from the time-series image data using the two line sensor cameras.

<Embodiment 3> Embodiments of a velocity measuring method and a pattern recognition method according to a third embodiment of the present invention, and an apparatus thereof will be described below. In the present embodiment, as an example of a plurality, the observation points where two line sensor cameras are installed are fixed, and velocity measurement and pattern recognition of a moving object are performed. The same measurement system as that of the second embodiment is used in this embodiment (see FIGS. 7 and 8).

FIG. 15 is a flowchart of the speed measuring method and the pattern recognizing method of this embodiment. Steps 1501 to 1505 shown in FIG. 15 are processes executed by the information processing apparatus 101 such as the CPU of a personal computer. The speed measurement and pattern recognition method of this embodiment will be described below with reference to the flowchart of FIG.

First, the moving speed of the moving object is obtained. First, an image change point on the time axis is detected from each of the time-series image data 1511 and 1512 (step 1501). However, the time-series image data is acquired as in the second embodiment. FIG. 16 is an example of a diagram showing temporal changes in time-series image data, in which 1601 is time-series image data A of the line sensor camera A, 1602 is a 1-line image and a 1-line image immediately before it. Change in the amount of change over time, 16
03 is an image change point, 1604 is a threshold value for the change amount, 1
Reference numeral 605 is a time axis. In this embodiment, since the observation point is fixed, there is almost no temporal change in the image when there is no moving object. On the other hand, the change in the image at the moment when the moving object starts to pass becomes large. Therefore, the temporal change 1602 of the image change amount is obtained by the difference calculation or the correlation coefficient calculation (see the reference (3)). If the number of pixels per line is m, the brightness (gradation) of the i-th (1 ≦ i ≦ m) pixel in the line at time t is I (t, i), and the line scanning time interval is t0, The above-mentioned time change 1602 is given as e (t) defined by the following equation.

[0067]

[Equation 3]

By the processing (threshold processing) based on the judgment as to whether or not the amount of change is greater than or equal to the threshold 1604, the image change point 1603
Can be asked. Similar processing is performed on the time-series image data B1512 of the line sensor camera B to obtain the image change point.

Next, the correspondence between each image change point and the time shift amount are calculated. When the moving direction of the moving vehicle is determined as in the present embodiment, the line sensor camera A is crossed and then the line sensor camera B is crossed. Therefore, the image change point is detected from the time-series image data A, and the image change point of the time-series image data B immediately after that is detected as the start point of the image pattern of the moving vehicle. Even when the moving direction of the vehicle is unknown, the neighboring image change points are associated as the start point or the end point of the image pattern of the moving vehicle. When the association is completed, the speed of the moving object can be calculated from the moving time and the actual moving distance as in the second embodiment.

The pattern recognition processing from step 1504 to step 1505 is obtained in the same manner as in the second embodiment.

<Embodiment 4> An embodiment of a velocity measuring method and a length measuring method according to a fourth embodiment of the present invention, and an apparatus for them will be described below. In this embodiment, as an example of a plurality, the observation points where two line sensors are installed are fixed, and the velocity and length of a moving object are measured. The measurement system of this embodiment is the same as that of the second and third embodiments (see FIGS. 7 and 8).

FIG. 17 is a flow chart of the speed measurement and length measurement method of this embodiment, and steps 1701 to 1704 shown in FIG. 17 are processes executed by the information processing apparatus 101 such as the CPU of the personal computer. Hereinafter, the speed measurement and length measurement method of this embodiment will be described with reference to the flowchart of FIG. First, the moving speed of the moving object is obtained. First, similarly to the second embodiment, the start point and the end point of the image change area with the background image are detected on the time axis from the time series image data 1711 and 1713 (step 1).
701).

FIG. 18 is an example of a diagram showing a temporal change in the amount of change between the time series image data and the background image.
Is the time series image data A of the line sensor camera A, 180
2 is a background image A, 1803 is a temporal change in the amount of change from the background image, 1804 is an image change point (start point), 1805 is an image change point (end point), 1806 is a threshold value for the change amount, 1
807 is a time axis. In this embodiment, since the observation point is fixed, there is almost no temporal change between the time-series image data and the background image when there is no moving object. On the other hand, the change in the image when the moving object is passing becomes large.
Therefore, the temporal change 1803 of the amount of change from the background image is obtained by difference calculation or correlation coefficient calculation (see reference (3)). The image change point (start point) 1804 and the image change point (end point) are processed by the threshold value processing with the threshold value 1806 for the change amount.
1805 can be obtained. Line sensor camera B
The same process is performed on the time-series image data B1713 to obtain the start point and the end point of the image change point.

Next, the correspondence between the image change points and the time shift amount are calculated. Here, the start point and the end point of the image change point are considered as an image change point set. When the moving direction of the moving vehicle is determined as in the present embodiment, the line sensor camera A is crossed and then the line sensor camera B is crossed. Therefore,
An image change point set is detected from the time-series image data A, and the image change point set of the time-series image data B immediately after that is associated as the start point and the end point of the image pattern of the moving vehicle. Even if the moving direction of the vehicle is unknown, a set of image change points in the vicinity is associated as a set of the start point and the end point of the image pattern of the moving vehicle. When the association is completed, the speed of the moving object can be calculated from the moving time and the actual moving distance as in the first embodiment.

In the length measuring process in step 1704, the time between the start point and the end point is equivalent to the photographing time of the moving object.
This can be performed in the same manner as in the second embodiment.

In the second, third, and fourth embodiments described above, the image on the side surface is targeted, but as shown in FIGS. 19 and 20, the line sensor cameras A1902 and 2003 with the moving vehicles 1901 and 2001 installed on the upper side and the line sensor cameras A1902 and 2003 are used. Sensor camera B19
It is also applicable to images taken from 03.2004. Further, as shown in FIGS. 21 and 22, since the processing results are accumulated, the line sensor cameras A2102 and 2202 are
The line sensor cameras B2103 and 2203 and the area sensors 2104 and 2204 may be synchronized with each other to display or record the result of pattern recognition on a still image or a video image of a moving object (moving vehicles 2101 and 2201). . Moreover, although two line sensor cameras are used, two or more line sensor cameras may be installed.

<Embodiment 5> Next, an embodiment of a vehicle system for controlling speed violation using the speed measuring method of the present invention will be described. In this embodiment, the observation point where two line sensor cameras are installed is fixed, and the speed measurement and the license plate recognition of the moving vehicle are performed.

FIG. 23 is a diagram showing an example of a speeding vehicle control system according to this embodiment. The line sensor camera A3102, the line sensor camera B3103, and the infrared illumination 3104 are installed on the gantry so that the moving vehicle 3101 can be photographed from above. The line axis of the line sensor camera A3102 and the line axis of the line sensor camera B3103 are set parallel to each other, and images are taken in a fixed line imaging cycle time in synchronization. In addition, the line sensor camera A310
2 and the line sensor camera B3103 are equipped with a polarizing filter for suppressing diffused reflection on the windshield to photograph the driver and a visible light cut filter (infrared transmission filter) for obtaining a stable image regardless of day or night. .

FIG. 24 is a side view of the system of FIG. 23. The line sensor camera A3203, the line sensor camera B3204, and the infrared illumination 3208 are installed at an angle θ downward with respect to the ground. At this time, the moving distance Lr of the moving vehicle between the line sensor cameras A and B is Lr = L / sin (θ).

FIG. 25 is an example of a time-series image acquired by the line sensor camera A and the line sensor camera B. In this way, by installing the line sensor camera from the upper side downward, it is possible to photograph the license plate and the driver.

FIG. 26 is a flow chart showing the processing of the speed violation vehicle control system using the speed measuring method of the present invention, and the outline according to this flow chart will be described below. First, the time series image data A of the line sensor camera A and the time series image data B of the line sensor camera B are acquired, and the image change point on the time axis is detected from each time series image data (step 3401).

FIG. 27 shows an example of a temporal change in the amount of change between a 1-line image and a 1-line image immediately before it.
The image change point can be detected by the threshold value for the change amount.

Next, the correspondence between the image change points between the time-series image data A and B and the movement time are calculated (step 3402). When the moving direction of the moving vehicle is predetermined as shown in FIG. 24, the line sensor camera A and the line sensor camera B are traversed in this order. Therefore, the image change point is detected from the time-series image data A, and the image change point of the time-series image data B immediately after that is associated as the starting point of the image pattern of the moving vehicle. Therefore, the moving time Tv with respect to the moving distance Lr between the two line sensor cameras of the moving vehicle
Is as shown in FIG. Next, the moving speed Vm of the moving vehicle is calculated using the moving distance Lr and the moving time Tv (step 3403). The moving speed Vm is calculated by the following equation. Vm = Lr / Tv

Next, the time-series image data and the template of the previously stored character pattern are corrected using the moving speed to calculate the degree of similarity (step 3404). And
Characters are recognized based on this similarity and a license plate recognition result is output (step 3405).

FIG. 29 shows an example of time-series image data of the license plate portion when the speed of the moving vehicle is slow and when it is fast. As shown in FIG. 29, the characters on the license plate are elongated when the speed is slow and are crushed when the speed is fast. Therefore, the height of the characters on the license plate can be normalized by dividing the time axis of each time series image data in FIG. 25 by the moving speed. If it can be normalized, it does not depend on the moving speed,
Conventional license plate recognition technology based on similarity ("Ch
aracter Recognition in Scene Images "(Reference (4), So
ciety of Manufacturing Engineers, 1989)) to identify the license plate of a moving vehicle. Here, since there are two types of image data to be recognized, one for the line sensor camera A and one for the line sensor camera B, the recognition accuracy can be improved by recognizing the license plate for each image. I can.

Further, if the license plate can be recognized, it is possible to find a completely coincident portion between the two images of the time-series image data A and B. Therefore, the moving time of the plate position is recalculated to improve the accuracy of the moving speed. Can be increased.

In this way, the speed of the moving vehicle above the legal speed can be measured, the characters on the license plate of the vehicle can be recognized, and the image of the driver can be stored.

In the above description, reference numerals 108, 205, 303, 702, 803, 120 assigned to the drawings.
2,1303,1902,2003,2102,220
Reference numerals 2 and 3102 represent the same line sensor camera A. Similarly, reference numerals 109, 206, 304, 70
3,804,1203,1304,1903,200
4, 2103, 2203, 3103, and 3204 represent the same line sensor camera B. Similarly, reference numerals 701, 801, 1201, 1301, 1901, 2
001, 2101, and 2012 each represent a moving vehicle that is a measurement target.

Similarly, reference numerals 306, 806 and 130
5, 2005 and 3205 each represent a line-axis distance L of the line sensor camera. Similarly, reference numeral 30
7, 807, 2007 and 3207 all represent the moving distance Lr of the measurement target. Similarly, reference numerals 506 and 11
03, 1403, and 3604 each represent a travel time Tv required to travel the travel distance Lr. Similarly, reference numeral 1
Reference numerals 104 and 1404 each represent a shooting time Tm from the start point to the end point.

The program for realizing the functions of the above processes is recorded in a computer-readable recording medium, and the program recorded in this recording medium is read into the computer system and executed to measure the speed. Pattern recognition or length measurement may be performed.
The “computer system” here means an OS.
And peripheral equipment.

The term "computer-readable recording medium" means a portable medium such as a floppy (registered trademark) disk, a magneto-optical disk, a ROM, a CD-ROM, or a storage device such as a hard disk built in a computer system. Say that. Further, the "computer-readable recording medium" is a volatile memory (RAM) inside a computer system which serves as a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those that hold the program for a certain period of time are also included.

The above program may be transmitted from a computer system that stores the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.

Further, the above-mentioned program may be for realizing a part of the above-mentioned functions. Further, it may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

Although the embodiments of the present invention have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments, and a design etc. within the scope not departing from the gist of the present invention are also possible. included.

[0095]

According to the present invention, an object thereof is a method for measuring a relative moving speed when a target object moves on a predetermined orbit relative to an observation point, and a line axis The time-series image data captured time-sequentially at a constant line imaging cycle time is acquired by synchronizing a plurality of line image acquisition devices installed so as to be parallel to each other and traverse the trajectory of the target object that moves relative to each other. And calculating the degree of similarity, by associating the image of the target object between the time-series image data, from the amount of deviation of the target object between the associated images and the line imaging cycle time. Relative between the target object and the observation point from the step of obtaining the movement time of the target object between each line image acquisition device, the movement time, and the distance between the image acquisition positions of each line image acquisition device on the trajectory Achieved by an image processing method having a step of obtaining the dynamic speed. By such a method, the line image acquisition device acquires a clear image at high speed,
By the calculation based on the image, it becomes possible to measure the relative moving speed of the target object with respect to the observation point.

Another object of the present invention is a method of recognizing a target object having a specific shape pattern recognizable as an image when the target object moves on a predetermined orbit relative to an observation point. That is, the step of obtaining the relative moving speed between the observation point and the target object using the speed measuring method, and by correcting the scale in the time axis direction based on the relative moving speed, time series image data And the step of matching the scale of the template of the specific shape pattern prepared in advance, and by calculating the similarity between the time-series image data and the template of the specific shape pattern with the same scale, And a step of detecting and recognizing the specific shape pattern in the series image data. By such a method, it becomes possible to easily recognize the pattern of the moving target object.

Further, according to the present invention, there is provided a method for measuring the length of a target object from each time-series image data obtained by photographing the target object moving on a predetermined orbit with a fixed observation point, A step of obtaining the moving speed of the target object using a speed measurement method, by determining whether the difference between the background image acquired in advance is a threshold value or more, the starting point of the target object from each of the time-series image data and Detecting an end point, obtaining a shooting time from the start point to the end point of the target object based on the line shooting cycle time, and calculating a length of the target object from the shooting time and the moving speed of the target object. With the image processing method including the step of performing, it is possible to measure the length of the target object.

Further, in the present invention, when the observation point is fixed and the target object moves on a predetermined orbit, the moving speed measurement, the target object length measurement, and the target object pattern recognition processing can be performed. become able to.

Further, in the present invention, when the line image acquisition device (line sensor camera or the like) is used, it is not necessary to install the device so as to sandwich the target object, and therefore the installation is easier than the phototube. Unlike the case of the phototube, since the target object can be stored as image data, the target object can be specified after the measurement.

Moreover, since the line image acquisition device enables high-speed photographing about 1000 times faster than that of a video camera, the speed of the object can be measured with higher accuracy than that of a conventionally used video camera, and an object moving at a high speed can be obtained. It is also possible to measure the speed of. Also, unlike when using a video camera,
By installing a plurality of line image acquisition devices in parallel, speed measurement is possible without depending on the distance from the camera.

Further, in the present invention, by increasing the number of line image acquisition devices used, speed measurement and length measurement can be performed by the number of combinations of line image acquisition devices, so that the measurement accuracy can be improved. . Further, even if the target object cannot be recognized from one image due to the influence of noise or the like, the recognition processing is applied to other images, so that the accuracy of pattern recognition can be improved.

Further, since the apparatus of the present invention can photograph only visible light, it is possible to measure the speed and the length and recognize the pattern without being perceived by the measuring object.

[Brief description of drawings]

FIG. 1 shows an example of a measuring device according to a first embodiment of the present invention.

FIG. 2 shows an example of a measurement system for recognizing a stopped vehicle according to the first embodiment.

FIG. 3 is a view of the measurement system of FIG. 2 seen from above.

FIG. 4 is a flowchart of a velocity measuring method and a pattern recognizing method according to the first embodiment.

FIG. 5 shows an example of output from the line sensor camera according to the first embodiment.

FIG. 6 is a diagram showing an example of a template of a specific pattern according to the first embodiment.

FIG. 7 shows an example of a measurement system when measuring a moving vehicle according to a second embodiment of the present invention.

8 is a diagram of the measurement system of FIG. 7 viewed from above.

FIG. 9 is a flowchart of a velocity measuring method, a length measuring method, and a pattern recognition method according to the second embodiment.

FIG. 10 shows an example of output from the line sensor camera according to the second embodiment.

FIG. 11 shows an example of an image pattern extraction result according to the second embodiment.

FIG. 12 is a diagram showing an example of a measurement system when measuring a moving vehicle according to a second embodiment, as viewed from above.

FIG. 13 is a diagram of an example of a measurement system when measuring a moving vehicle according to the second embodiment, as viewed from above.

FIG. 14 shows an example of output from the line sensor camera according to the second embodiment.

FIG. 15 is a flowchart of a velocity measurement and pattern recognition method according to a third embodiment of the present invention.

FIG. 16 shows an example of a graph of an output from a line sensor camera and an image change amount according to the third embodiment.

FIG. 17 is a flowchart of a velocity measurement and length measurement method according to the fourth embodiment of the present invention.

FIG. 18 shows an example of a graph of the output from the line sensor camera and the image change amount with respect to the background image in the fourth embodiment.

FIG. 19 shows an example of another measurement system for measuring the moving vehicle according to the second, third, and fourth embodiments of the present invention.

20 is a side view of the measurement system of FIG.

FIG. 21 shows an example of a measurement system in the case of measuring the moving vehicle of Embodiments 2, 3, and 4 of the present invention and recording the measurement result in synchronization with the output of the area sensor.

FIG. 22 shows an example of a measurement system for measuring a moving vehicle according to Embodiments 2, 3, and 4 of the present invention and recording the measurement result in synchronization with the output of the area sensor.

FIG. 23 shows an installation situation of a moving vehicle speed measurement and license plate recognition device according to a fifth embodiment of the present invention.

FIG. 24 is a side view showing an example of measuring the speed of a moving vehicle according to a fifth embodiment of the present invention.

FIG. 25 shows an example of an output from the line sensor camera according to the fifth embodiment of the present invention.

FIG. 26 is a flowchart of a method for speed measurement and license plate recognition of a moving vehicle according to Embodiment 5 of the present invention.

FIG. 27 shows an example of a graph of the output from the line sensor camera and the image change amount in the fifth embodiment.

FIG. 28 shows an example of an image pattern extraction result in the fifth embodiment.

FIG. 29 is a diagram showing speed correction of a license plate image according to the fifth embodiment.

[Explanation of symbols]

101 information processing device 102 bus line 103 storage device 104 display device 105 measurement device 106 measurement value storage unit 107 image storage unit 108 line sensor camera A 109 line sensor camera B 110 synchronization device 201, 202 stopped vehicle 203 moving vehicle 204 moving direction 205 Line sensor camera A 206 Line sensor camera B 301 Moving vehicle 302 Moving direction 303 Line sensor camera A 304 Line sensor camera B 305 Angle θ 306 Line axis distance L 307 Moving distance Lr 308, 309 Stop vehicle 501 When line sensor camera A Series image data A 502 Time series image data of line sensor camera B 503 Strip image 504 Matched area 505 Strip width W 506 Moving time Tv 507 Time axis 601 Template 604 of specific pattern Corrected specific pattern template 701 Moving vehicle 702 Line sensor camera A 703 Line sensor camera B 704 Measuring device 801 Moving vehicle 802 Moving direction 803 Line sensor camera A 804 Line sensor camera B 805 Angle θ 806 Line axis distance L 807 Moving Distance Lr 1001 Time-series image data A 1002 of the line sensor camera A 1002 Time-series image data B 1003 of the line sensor camera B Background image A 1004 Background image B 1005 Time axis 1101 Image pattern A 1102 Image pattern B 1103 Travel time Tv 1104 Shooting time Tm 1105 Time axis 1201 Moving vehicle 1202 Line sensor camera A 1203 Line sensor camera B 1301 Moving vehicle 1302 Moving direction 1303 Line sensor camera A 1304 line Sensor camera B 1305 Line distance L 1401 Time-series image data A 1402 of line sensor camera A 1402 Time-series image data of line sensor camera B B 1403 Moving time Tv 1404 Shooting time Tm 1405 Time axis 1601 Time series of line sensor camera A Image data A 1602 Time change of change amount between one line image and one line image immediately before 1603 Image change point 1604 Threshold for change amount 1605 Time axis 1801 Time series image data A 1802 of line sensor camera A Background image A 1803 Change in amount of change with background image 1804 Image change point (start point) 1805 Image change point (end point) 1806 Threshold for change amount 1807 Time axis 1901 Moving vehicle 1902 Line sensor camera A 1903 Line sensor camera B 2001 Moving vehicle 2002 moving direction 2003 line sensor camera A 2004 line sensor camera B 2005 line distance L 2006 angle θ 2007 moving distance Lr 2101 moving vehicle 2102 line sensor camera A 2103 line sensor camera B 2104 area sensor 2201 moving vehicle 2202 line sensor camera A 2203 Line sensor camera B 2204 Area sensor 3101 Moving vehicle 3102 Line sensor camera A 3103 Line sensor camera B 3104 Infrared illumination 3105 License plate 3201 Moving vehicle 3202 Moving direction 3203 Line sensor camera A 3204 Line sensor camera B 3205 Line distance L 3206 Angle θ 3207 Moving distance Lr 3208 Infrared illumination 3301 Time series image data of line sensor camera A A 3302 Time-series image data of line sensor camera B 3303 Time axis 3501 Time-series image data of line sensor camera A A 3502 Temporal change in change amount between 1-line image and 1-line image immediately before it 3503 Image change point 3504 Threshold for change amount 3505 Time axis 3601 Time series image data of line sensor camera A 3602 Time series image data of line sensor camera B 3603 Time axis 3604 Movement time Tv 3701 Time axis 3702 Time series image data of license plate (speed is Slow) 3703 Time series image data of license plate (when speed is high) 3704 License plate image normalized based on speed

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Sakuichi Otsuka 2-3-1, Otemachi, Chiyoda-ku, Tokyo Nippon Telegraph and Telephone Corporation (56) Reference JP 2000-162220 (JP, A) JP 2000-97968 (JP, A) JP-A-6-66820 (JP, A) JP-A-4-198870 (JP, A) Actual development Sho-59-156673 (JP, U) US Patent 4135817 (US, A) International Publication 98/054671 (WO, A1) (58) Fields investigated (Int.Cl. ⁷ , DB name) G01B 11 / 00-11 / 30 G01P 3/68

Claims (6)

(57) [Claims] 1. An image processing method for recognizing a target object having a specific shape pattern recognizable as an image when the target object moves on a predetermined orbit relative to an observation point, The trajectory of the target object whose line axes are parallel and move relative to each other
Tilt at a specified angle to the relative movement
Fixed by synchronizing multiple line image acquisition devices installed
Each time series taken in time series with the line shooting cycle time of
By obtaining the image data and calculating the similarity, the time series image data
The images of the target object are associated with each other, and the associated images
Between the amount of displacement of the target object and the line imaging cycle time
The moving time of the target object between each line image acquisition device
Step, the travel time, and acquisition of each line image on the orbit
A step of obtaining a relative movement speed between the observation point and the target object from the distance between the image acquisition positions of the apparatus and the predetermined angle; and a time axis of the time-series image data with the relative movement speed . Percent
By correcting the scale in the time axis direction,
Calculating the similarity between the time-series image data in advance and the step of providing been matched the scale of the template of the specific shape pattern, template matching the specific shape pattern and the time-series image data of the scale Accordingly, there is a step of recognizing the specific shape pattern in the time-series image data. 2. An image processing method for recognizing a target object having a specific shape pattern recognizable as an image when the target object moves on a predetermined trajectory, wherein line axes are parallel and move. Traverse the trajectory of the target object
It is installed so that it cuts at a predetermined angle to the moving direction.
Multiple line image acquisition devices are synchronized to maintain a constant line.
Each time-series image data taken in time-series with the shooting cycle time
Data acquisition step and calculating the similarity between the time-series image data.
, The object between the time series image data on the time axis
Correspond to the body, or the background image acquired in advance with each of the time series image data
An image change point whose difference from the image is equal to or greater than a threshold value is detected, and
Based on the image change points in each time series image data,
The target object is paired between the time series image data on the axis.
Responsive or each line in each time series image data
If the difference between the image and the line image each one cycle before is greater than or equal to the threshold value,
Image change points are detected and added to the time series image data.
Based on the image change point, the time series images are displayed on the time axis.
By associating the target object between the image data, the image acquisition positions of the line image acquisition devices of the target object
The amount of deviation in the line shooting cycle is calculated.
Obtaining a moving time from the time, obtaining the moving time and each line image on the orbit
A step of obtaining a moving speed of the target object from the distance between the image acquisition positions of the apparatus and the predetermined angle; a time axis of the time-series image data is divided by the moving speed ;
By performing the time-axis direction of the scale of the correction, the step to match the scale of the template of the specific shape pattern prepared in advance and the <br/> time-series image data, the time-series image data that matches the scale And a step of recognizing the specific shape pattern in the time-series image data by calculating a similarity between the specific shape pattern template and the template of the specific shape pattern. 3. An image processing apparatus which recognizes a target object including a specific shape pattern recognizable as an image when the target object moves on a predetermined orbit relative to an observation point, The trajectory of the target object whose line axes are parallel and move relative to each other
Tilt at a specified angle to the relative movement
Synchronize multiple installed line image acquisition devices.
The shooting operation is performed in time series with a fixed line shooting cycle time, and
An image acquisition unit that acquires time-series image data, and the time-series image data are calculated by calculating the degree of similarity.
The images of the target object are associated with each other, and the associated images
Between the amount of displacement of the target object and the line imaging cycle time
The moving time of the target object between each line image acquisition device
Time calculation means , the moving time, and acquisition of each line image on the orbit
From the distance between the image acquisition positions of the device and the predetermined angle
Speed calculation means for obtaining a relative movement speed between the target object and the observation point, and time of the time-series image data acquired by the image acquisition means
Split between shaft by the relative movement speed determined by the speed calculating means, by correcting the scale of the time axis direction, the specific shape in which the image acquisition means is prepared in advance and the acquired time-series image data Detecting the specific shape pattern by calculating a similarity between the scale correction unit that matches the scale of the pattern template and the template of the time-series image data that matches the scale and the specific shape pattern. An image processing apparatus comprising: a pattern detecting unit for recognizing. 4. An apparatus for recognizing a target object at a stationary observation point when the target object including a specific shape pattern recognizable as an image moves on a predetermined orbit, the line axes being parallel. And traverse the trajectory of the moving target object.
It is installed so that it cuts at a predetermined angle to the moving direction.
Multiple line image acquisition devices
Each time-series image
By calculating the degree of similarity between the image acquisition means for acquiring image data and each of the time-series image data,
, The object between the time series image data on the time axis
Correspond to the body, or the background image acquired in advance with each of the time series image data
An image change point whose difference from the image is equal to or greater than a threshold value is detected, and
Based on the image change points in each time series image data,
The target object is paired between the time series image data on the axis.
Responsive or each line in each time series image data
If the difference between the image and the line image each one cycle before is greater than or equal to the threshold value,
Image change points are detected and added to the time series image data.
Based on the image change point, the time series images are displayed on the time axis.
By associating the target object between the image data, the image acquisition positions of the line image acquisition devices of the target object
The amount of deviation in the line shooting cycle is calculated.
Time calculating means for calculating a travel time from time, the travel time, and acquisition of each line image on the orbit
From the distance between the image acquisition positions of the device and the predetermined angle
A speed calculation means for obtaining the moving speed of the target object, and a time calculation means for calculating the moving speed of the target object
Split between shaft by said movement speed obtained by said speed calculating means, by correcting the scale of the time axis direction, the specific shape pattern by the image acquisition means is prepared in advance and the acquired time-series image data The scale correction means for matching the scale with the template and the similarity between the time-series image data with the matched scale and the template for the specific shape pattern are detected to recognize and recognize the specific shape pattern. An image processing apparatus, comprising: 5. A computer program for executing the steps of the image processing method according to claim 1 or 2 in a computer. 6. The method of claim 1 or 2 computer readable recording medium recording a computer program for executing the steps of the image processing method in a computer according to.