JP3849420B2 - Interface measuring device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an interface measurement apparatus including an image acquisition unit for acquiring an original image including an interface to be measured, and an image analysis unit that analyzes the acquired original image to obtain an interface position.
[0002]
[Prior art]
The interface refers to a boundary where two layers such as a liquid, a gas, and a solid are in contact, for example, a boundary between the liquid and air in a drum in which the liquid is stored. The following are examples of commercially available apparatuses for measuring the interface.
(1) Vertical reflection displacement meter
According to this principle, laser light or ultrasonic waves are emitted vertically downward from the liquid surface toward the liquid surface, and the reflection on the liquid surface is received to measure the distance.
(2) Capacitance type level switch
This principle detects whether there is a liquid level in the vicinity of the sensor using a difference in relative permittivity between liquid and air.
[0003]
However, in the case of the above (1), since the device needs to be installed in one of the two layers, a device that does not affect the device, for example, the device can be in a shielded state. Necessary. In addition, when it is necessary to be closed from the outside such as in a tank, there is a problem that installation and maintenance are not easy.
[0004]
In the case of (2), detection at the position where the sensor is provided is possible, but there is a problem that continuous values cannot be measured over a wide range.
In addition to the above, there are also float level meters, resistance level meters, and impedance type interface meters, but these have large differences in characteristic values such as impedance and density at the interface, and the values are stable. The precondition to say was needed. In actual site measurement, there are various instability factors, and it is often difficult to apply the above-mentioned commercially available apparatus as they are.
[0005]
As a prior art considering the above problems, there is a water level measuring device by image processing disclosed in Japanese Patent Laid-Open No. 8-43173.
This device captures an image of a water level gauge that measures the water surface position (interface) of the liquid, automatically measures the water level by image processing, converts the image of the water level meter into a gray image, and the gray value thereof Are obtained in a horizontal direction, and a histogram indicating a change in the vertical direction is obtained. When the histogram changes beyond a predetermined threshold value set in advance, the position is obtained as a water level.
Other examples of the interface measuring device using image processing include the devices disclosed in Japanese Utility Model Laid-Open Nos. 6-30727 and 7-98238.
[0006]
[Problems to be solved by the invention]
However, these conventional interface processing apparatuses based on image processing have a problem that they cannot be used stably in an on-site environment with many disturbances.
In other words, in the conventional image processing, there is a change in some physical quantity (density, etc .: change in the histogram in the above example) at the interface, and under the premise that the same change does not occur in any part other than the interface. The interface can be detected stably.
[0007]
On the other hand, when the detection is performed based on the density difference between the two layers forming the interface, the density difference may not be sufficient because each density is almost determined by the physical properties of the two layers forming the interface. Although there is a method for producing a concentration difference by coloring the substance forming the two layers, it is often difficult to improve the concentration difference at the interface due to restrictions on the apparatus. That is, it is often necessary to detect the interface based on an image having a very small density difference.
[0008]
In addition, the actual interface measurement site is often outdoors, in which case there are various disturbances such as dirt on the observation surface, changes in ambient brightness, and reflections of shadows from surrounding objects. Captured as an image. Moreover, the concentration change due to these disturbances is often larger than the concentration change at the interface to be measured. In such a case, it is necessary to extract only the change in concentration at the interface in distinction from the change due to other disturbances.
[0009]
The present invention has been made in view of the above circumstances, and an object of the present invention is to provide an interface measuring apparatus that can accurately measure an interface even when a disturbance occurs.
[0010]
[Means for Solving the Problems]
In order to achieve the above object, an interface measurement apparatus according to the present invention includes an image acquisition means for acquiring an original image including an interface to be measured, and an image analysis means for analyzing the acquired original image to obtain an interface position. In the interface measuring apparatus, the image analysis means includes a difference calculation unit for obtaining a difference image between original images acquired at different times, and an extracted image obtained by extracting a change in the original image based on the difference image. With respect to a change extraction unit to be obtained, an interface position determination unit that determines an interface position based on the extracted image, and a change extraction image obtained in the process of interface measurement, the change extraction image depends on a change in the interface. And a change extraction image evaluation unit for determining whether or not the image is a thing.
[0011]
Explaining the principle of the present invention, it is intended to capture not only the concentration difference but also the temporal change in the position of the interface as a physical quantity that changes at the interface. For example, in the case of an interface in an operating production plant, the interface often oscillates constantly. On the other hand, disturbance images such as dirt on the observation surface, changes in ambient brightness, and shadows of surrounding objects may change when viewed over a long period of time, but can be considered to be stationary over a short period of time. Many.
[0012]
Therefore, an original image (an original image means an image before image processing) is acquired at different times, and a difference image between the current original image and a past original image is obtained by the difference calculation unit. . What is necessary is just to set suitably the time interval between the present original image and the past original image according to the period of rocking | fluctuation of an interface. By calculating the difference image, a disturbance image in a stationary state can be removed, and an extracted image that leaves only the oscillating interface image can be obtained. That is, the interface is not detected based on the magnitude of the density change, but the change in the interface position is captured. Even if the density change at the interface is smaller than the density change of the disturbance image, it can be detected. The interface position determination unit finally determines (specifies) the interface position based on the extracted image.
As described above, it has been possible to provide an interface measuring apparatus that can accurately measure an interface even when a disturbance occurs.
[0013]
As a preferred embodiment of the present invention, the image analysis means analyzes a relative positional relationship between two original images acquired at different times, corrects a positional deviation between the two original images, and then calculates a difference. And a misalignment correcting unit that takes the position. In an actual field environment, a large-scale apparatus is often adjacent to each other, and there is often a difference between the original images acquired at different times over the entire image due to vibrations from these adjacent apparatuses. When such a difference is taken between original images, a noise component remains over the entire image. In such a case, it is preferable to correct the misalignment before taking the difference between the original images, and shift the one image in the reverse direction by the amount of the deviation before taking the difference. For the analysis of the positional relationship, a known method such as pattern matching or normalized correlation can be used. Thereby, the noise component which appears in a difference image can be reduced, and interface measurement can be performed accurately.
[0014]
As another preferred embodiment of the present invention, the image analysis means further includes a smoothing processing unit that obtains a smoothed image obtained by removing noise components from the difference image. Even if the difference image is obtained by taking the difference between the sampling images, it is normal that the disturbance image cannot be completely removed, and some noise components remain. When the noise component is small, it is not necessary to perform the smoothing process, but when the noise component is conspicuous, it is preferable to smooth the difference image to obtain a smoothed image. Thereby, interface measurement can be performed with high accuracy.
[0015]
As another preferred embodiment of the present invention, the image processing apparatus further includes a threshold value calculation unit that obtains a threshold value for obtaining the extracted image from the density histogram of the difference image or the density histogram of the smoothed image. Can be given.
[0016]
When the brightness of the image changes due to the influence of the day and night weather, the density histogram also changes. Therefore, the extracted image (for example, 0 (black) level, 1 (white) level, gray label 3) It is preferable to obtain a threshold value for obtaining a value image. Thereby, interface measurement can be performed with high accuracy. Note that the threshold value may be fixed when there is no influence of daytime and nighttime weather and the brightness of the image is stable (the density histogram does not change).
[0017]
As yet another preferred embodiment of the present invention, the change extraction unit is configured to use a threshold value obtained by the threshold value calculation unit or a preset threshold value from the difference image or the smoothed image. The thing which calculates | requires the said extracted image is mention | raise | lifted.
[0018]
In the difference image or the smoothed image obtained from the difference image, the pixel density corresponding to the change is estimated to be within a certain range, so the image of the change is extracted by setting a threshold value. can do.
[0019]
As still another preferred embodiment of the present invention, the interface position determination unit determines an interface position by comparing a density change pattern obtained from the difference image with a density change pattern prepared in advance. Things can be raised.
[0020]
When the extracted image is obtained, it is necessary to specify the interface position from this image. At that time, a density change pattern is obtained from the difference image and is compared with a density change pattern prepared in advance. As this density change pattern, for example, several possible patterns may be prepared, for example, when the interface moves upward or when the interface moves downward. Based on these patterns, the interface position can be finally identified from the extracted image.
[0021]
As still another preferred embodiment of the present invention, the image processing apparatus includes a display unit that displays the original image acquired by the image acquisition unit, and the display unit determines the interface position and the interface position determined by the interface position determination unit. One that displays a graph representing a change over time together with the original image is available.
[0022]
According to this configuration, the interface position and the temporal change of the interface position can be displayed together with the original image on the display means, so that the operator can immediately recognize the measurement result. Further, by displaying a graph showing the change in the interface position over time, not only the current position of the interface position but also the past measurement results can be recognized immediately.
[0023]
As yet another preferred embodiment of the present invention, a plurality of past original images for obtaining a difference image with respect to an original image to be subjected to interface measurement are prepared, and images for each past original image are prepared. An analysis is performed, and the interface position is determined by adopting the most preferable image analysis result from the respective image analysis results.
[0024]
According to this configuration, when there is almost no change in the position of the interface between the two original images, a component due to the change in the interface may be hardly extracted in the difference image. In such a case, measurement is impossible. By preparing a plurality of target images to be taken, a possibility that a component due to a change in the interface is extracted from a difference image from any one of the images increases, and measurement with higher reliability is possible. At this time, for example, it is effective to select an image when the interface measurement value is the maximum and an image when the interface measurement value is the minimum in the past predetermined time width as the target images for which the difference is taken.
[0025]
When specifying the interface, compare the results of image analysis using multiple difference target images, adopt the one that has the highest possibility that the extracted component is due to the change in the interface, and based on this, select the interface Is preferably specified. (Hereinafter, the fact that the extracted component is likely to be due to a change in the interface will be referred to as “appears to be an interface”.) As a method for determining whether or not it is likely to be an interface, for example, the following method may be mentioned. .
[0026]
Usually, when receiving only the force of gravity, the interface is expected to be nearly horizontal. Therefore, if the extracted image components are analyzed for features, they are concentrated in a limited range in the vertical direction and distributed over a range close to the width of the observation window at the interface in the horizontal direction. Can be judged to be an interface. For this purpose, the vertical and horizontal dimensions (generally called the ferret diameter) of the circumscribed rectangle of the extracted image are evaluated, or the range in which the projected image is distributed and the maximum frequency are evaluated by taking a vertical projection of the extracted image. And the like. Further, whether or not the center of gravity of the extracted image is located near the center in the horizontal direction of the interface observation window is also a clue for evaluation.
[0027]
As a method for evaluating whether or not it is likely to be an interface, another preferred embodiment is a method using a neural network using the extracted image as input data. As described above, when the likelihood of the interface is determined from the feature amount of the extracted image, the value of the feature amount and the value of the interface property must be matched with each other, but this operation is not easy. Therefore, we built a neural network that uses the image pattern extracted as the interface candidate as input data, and learned it based on a lot of actual image analysis results and teacher data that made a decision as to whether or not a human is an interface in advance. It is good to determine the likelihood of the interface based on the learning result. Among the image analysis results for a plurality of difference target images, the one with the largest output value of this neural network is adopted as the most likely interface, and the position of the interface can be specified.
[0028]
In order to achieve the object of the present invention, an interface measuring method according to the present invention includes a step of acquiring an original image including an interface to be measured, and obtaining a difference image between the acquired current original image and a past original image. The method includes a step, a step of obtaining an extracted image obtained by extracting a change amount of the original image based on the difference image, and a step of determining an interface position based on the extracted image.
[0029]
The effect | action by this structure is the same as that of the interface measuring apparatus which concerns on this invention, and is as follows.
(B) Acquire an original image including the interface to be measured. The sampling interval may be set according to the interface fluctuation state.
(B) An extracted image obtained by extracting a change in the original image based on the difference image is obtained. Here, “based on the difference image” includes not only the case of extracting the change of the original image directly from the difference image, but also another image obtained from the difference image, for example, noise from the difference image. This includes a case where a smoothed image from which components are removed is obtained, and a change is extracted from the smoothed image.
(C) The interface position is determined based on the extracted image.
Since other operations and effects are the same as those of the interface measuring apparatus, description thereof will be omitted.
[0030]
DETAILED DESCRIPTION OF THE INVENTION
A preferred embodiment of an interface measuring apparatus according to the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing an overall configuration (site + measurement room) including an interface measurement device.
<Overall configuration>
In the field, a container (such as a drum) 1 as an interface measurement target, an illuminating means 2 such as a fluorescent lamp that illuminates the measurement interface 1b that can be seen from the observation window 1a of the container 1, and an image for acquiring the interface 1b. A TV camera 3 (corresponding to image acquisition means) is provided.
In the present embodiment, the site is outdoors, and it is unavoidable that dirt adheres to the observation window 1a. In addition, the influence may appear as a disturbance in the image when it rains. Even if there is such a disturbance, the interface position is detected by devising an algorithm for acquiring the interface image using the existing TV camera 3 and analyzing the image.
[0031]
The measurement room is provided with an image analysis system 4 to which an image is input from the TV camera 3, and a TV monitor 5 and a CR monitor 6 (each corresponding to display means) connected to the image analysis system 4. Yes. The image analysis system 4 will be described next, but an image analysis means 40 is provided.
[0032]
<Image analysis system>
Next, a block configuration of the image analysis system 4 is shown in FIG.
The image analysis system 4 includes a system control unit 41 that controls each unit, an image analysis unit 40 that analyzes an image acquired by the TV camera 3 and measures an interface position, a hard disk 42, a floppy disk 43, data A recording / reproducing unit 44, a peripheral processing unit 45, an image synthesis unit 46, a graph creation unit 47, and a keyboard 48 are provided.
[0033]
FIG. 3 is a block diagram illustrating the function of the image analysis means 40. FIG. 4 is a flowchart showing a schematic procedure. The analog image signal acquired from the TV camera 3 is taken into the image analysis unit 40 at a predetermined sampling interval (cycle) as a digital image signal (original image) converted into digital data by the A / D converter 10. The sampling interval may be set according to the period at which the interface swings, and in this embodiment, it is set to 2 seconds. The captured original image is captured in the frame memory unit 40a. The frame memory unit 40a includes at least a first frame memory FM1 and a second frame memory FM2. When a plurality of difference target images are prepared, the same number of frame memories are required. In this embodiment, since three difference target images are prepared, in addition to the above, frame memories FM3 and FM4 and a frame memory (not shown) are secured as work areas for image analysis. The original image means an image before image processing is performed, and is represented as an 8-bit grayscale image (256-level grayscale image data) in this embodiment.
[0034]
First, the original image data is taken into the first frame memory FM1 (see s1 in FIG. 4). This original image (current original image) is shown in FIG. Further, the original image acquired two seconds ago is stored in the second frame memory FM2, and this original image (past original image) is shown in FIG.
[0035]
The difference calculation unit 40b obtains a difference image between the current original image and the past original image (s2). In the present embodiment, the past original image is subtracted from the current original image, but of course the reverse may be performed. Note that the difference image is obtained by offsetting the subtraction result by 128 (intermediate gradation), and this offset difference image is shown in FIG. The difference image is stored in a frame memory secured as a work area.
[0036]
The difference calculation unit 40b analyzes the relative position between the two images based on the normalized correlation before obtaining a difference image between the current original image and the past original image, and the frame memory FM1 is as much as the positional deviation between the two images. This function has a function of correcting misalignment by shifting the image. However, in the present embodiment, the image shift due to vibration or the like in the field environment did not occur so as to affect the subsequent analysis.
[0037]
When the difference image is obtained, the non-moving portion of the disturbance image or the like becomes 0 (128 when offset) when the difference is taken. However, even if the difference is actually taken, a minute noise component may remain. In order to remove such noise components, the smoothing processing unit 40c performs a smoothing process (s3). As the smoothing process, smoothing is performed using a 1 × 32 filter in this embodiment. This is a moving average filter, and is a method in which an average of 32 pixels centered on a target pixel is used as a smoothed image at the target pixel position. In the present embodiment, the reason why the filter is employed is that the influence on the interface image is small and the most effective is obtained. Of course, the smoothing filter is not limited to the above-mentioned filter, and an appropriate filter may be adopted according to the field situation. Further, when there is almost no noise component in the difference image, the smoothing process is unnecessary.
[0038]
Next, a process for extracting a change in the original image from the smoothed difference image will be described. This change corresponds to an image near the interface. This is based on the idea that the interface can be measured by extracting the amount of change in the original image because the interface is considered to be constantly oscillating.
[0039]
First, the density histogram of the smoothed difference image shown in FIG. 7 is obtained as shown in FIG. 8, where (a) is the density histogram when there is no interface change, and (b) is the interface change. FIG. In the case of FIG. 8 (a), the threshold value calculation unit 40d has both ends n of the frequency distribution due to noise. ₁ And n ₂ The
Set as threshold. In the case of (b), the threshold value calculation unit 40d is n in the middle of the frequency distribution due to noise and the frequency distribution due to changes in the interface position. _Three And n _Four The threshold
Set as. Since the width of the frequency distribution varies each time due to weather, time, vibration, electrical noise, and the like, the threshold value calculation unit 40d calculates the threshold value by mathematically recognizing the shape of the frequency distribution every time.
When the density histogram of the difference image is almost constant each time, the process for automatically calculating the threshold value is not necessary, and a fixed value set in advance may be used. It is effective to perform the above processing when the shape of the density histogram varies every time.
[0040]
When the threshold value is obtained, the change extraction unit 40e converts the difference image of FIG. 7 into a binary image. Specifically, the concentration is n ₁ ~ N ₂ Or n _Three ~ N _Four Between
The image is converted to 0 level (0 value) and n ₂ Or n _Four The above image is one level (255
Value) and n ₁ Or n _Three The following image is converted to an intermediate gray level. this
Thus, it is possible to obtain an extracted image (ternary image) obtained by extracting the change of the original image. This extracted image is shown in FIG. When obtaining this extracted image, it is preferable to perform isolated point removal processing (expansion / contraction processing, etc.) for removing noise components.
[0041]
FIG. 9 is an ideal extracted image, and two islands appear. The two appear because the interface in the present embodiment has an emulsion layer corresponding to a model pattern shown in FIG. 13 described later, and an algorithm for specifying the correct interface position from this extracted image is required. This process is performed by the interface position determination unit 40f. Hereinafter, a procedure for specifying the interface position will be described.
[0042]
First, a density distribution pattern as shown in FIG. 10 is obtained from the difference image. This density distribution pattern is a cumulative average (or integration) of the cross-sections in the vertical direction of the difference image, and has a value of 128 in a portion where there is no change (straight line portion). FIG. 10A shows the density distribution pattern y and two islands (represented by a white level large ellipse E1 and a gray level small ellipse E2) in the extracted image of FIG. It is. The four lines L1, L2, L3, and L4 shown in the figure indicate positions where the density distribution pattern crosses the threshold value shown in FIG. In the example of FIG. 8B, the lines L3 and L4 have a threshold value n. _Four , The line L1, L2 is the threshold value n _Three The positions crossing each are shown. In terms of the relationship with the two islands, the large ellipse E1 is sandwiched between the lines L3 and L4, and the small ellipse E2 is sandwiched between the lines L1 and L2.
[0043]
In order to specify the interface position from this density distribution pattern, several density distribution patterns are prepared in advance. The density distribution pattern of the image near the interface varies depending on the object, but a model is created according to each density change pattern, and the appropriate model is selected from the density distribution pattern of the difference image obtained each time. The interface position can be determined.
[0044]
The model of FIG. 11 is a model in the case of a simple interface and the interface position moves upward with respect to the previous time. The density distribution pattern when the difference is taken is as shown on the right side of FIG. The interface position is above the convex portion of the density distribution pattern.
The model of FIG. 12 is a model in the case of a simple interface, and the interface position has moved downward relative to the previous time. The density distribution pattern when the difference is taken is as shown on the right side of FIG. The interface position is below the concave portion of the concentration distribution pattern.
[0045]
The model in FIG. 13 is a model in the case where there is a transition region such as an emulsion layer, and the interface position moves upward with respect to the previous time. The concentration distribution pattern in the case of taking the difference is the right side of FIG. The interface position is above the convex portion of the density distribution pattern.
The model in FIG. 13 is a model in the case where the meniscus layer has a density change and the interface position moves upward with respect to the previous time. The density distribution pattern in the case of taking the difference is shown on the right side of FIG. Thus, the interface position is above the convex portion of the density distribution pattern. Four typical models have been exemplified above.
[0046]
Returning to FIG. 10 (a), when the actually obtained density distribution pattern is compared with the model, it corresponds to the model of FIG. 13, and therefore the line L2 (the upper end of the white-level large ellipse E2) is taken as the interface position. decide. In the example of FIG. 10B, the line L1 (the lower end of the large ellipse E2) is determined as the interface position.
[0047]
The above is a series of image analysis algorithms. In this embodiment, as a more preferable embodiment, in addition to the original image acquired at the previous measurement as a target image for which a difference is to be obtained, Prepare an original image (FM3) when the position of the interface is the highest and an original image (FM4) when the interface position is the lowest, perform image analysis with these three original images, The image analysis result that seems to be the most interface was adopted from the inside, and the position of the interface was specified based on the image analysis result. It is desirable that the time width is longer than the time when the fluctuation of the interface stops irregularly and shorter than the time when the surrounding environment (such as the degree of contamination of the observation window) changes. In this embodiment, it is 30 minutes.
[0048]
In the present embodiment, the determination using the neural network is performed as a method for determining the likelihood of the interface. The network structure was a three-layer neural network with 50 points of input data, 50 points of intermediate layer, and 1 point of output. As shown in FIG. 17, the input data is obtained by dividing the interface candidate image extracted from each image into two blocks in the vertical direction around the interface candidate position and 25 blocks in the horizontal direction. Mosaic was made into such 50 points of data, and these 50 points were used as input data. The teacher data includes, for example, data as shown in FIG. 18 as a typical example. A large number of such teacher data are created from actual image analysis results, and learning is performed based on these teacher data.
[0049]
Of course, it is possible to use a neural network having a structure other than that of the present embodiment as a method for determining the likelihood of interface, and in reality, it is natural to determine it according to the shape of the measurement object actually observed.
[0050]
As a method for determining the likelihood of an interface, a method for determining from the feature amount of an image extracted as an interface candidate is also effective in addition to the present embodiment. (For example, the aspect ratio of the ferret diameter, whether the center of gravity is located in the center of the observation window, the major axis / minor axis ratio of the approximate ellipse, etc.)
[0051]
Returning to the flowchart of FIG. 4, when the process of specifying the interface position is completed (s3), the current original image data in the first frame memory FM1 is transferred to the second frame memory FM2 (s4). If the data is to be updated (s5), the data update process is performed (s6), and if it is not the data update timing, the process returns to s1.
[0052]
<Example of TV monitor display screen>
FIG. 15 shows a display screen example of the TV monitor 5. On this display screen, an original image obtained by synthesizing the line indicating the interface position, the trend graph, and the date and time is displayed. In FIG. 15, the interface position is visible in the middle on the right side, and the finally specified interface position is indicated by one horizontal line. A trend graph is a graph in which the horizontal axis represents time and the vertical axis represents measurement data. The measurement data refers to coordinate values in the interface position image obtained by each measurement. The vertical axis indicates the interface position. For example, the amount of displacement from the reference position is expressed in units of length.
[0053]
The trend graph is a graph of changes in measurement data in the past n hours. n can be arbitrarily set. For example, measurement data for 4, 8, and 12 hours can be displayed on a graph. Settings are input from the keyboard 48. In plant operation and the like, it is often important to know how the measurement data (interface position) has changed in the past several hours. In such cases, it is convenient to display a trend graph. The trend graph is automatically rewritten every predetermined time (for example, 1 minute).
[0054]
There is an operation indicator at the upper left of the trend graph on the display screen of the TV monitor 5, and the dot expands and contracts at each measurement, and the operator knows that the interface measurement system is operating normally. be able to. The date and time are displayed in the lower left of the display screen, but you may also display the name of the research department together. The date display position is preferably moved little by little as time passes to protect the screen.
[0055]
Returning to the block diagram of FIG. 2, the data recording / reproducing means 44 stores the measurement data in a predetermined file format for each date in the hard disk 42 which is an external storage device at a preset time interval. The data recording / reproducing means 44 also includes the measurement data stored in the hard disk 42 for the year / month / day (for one day) or year / month (for one month) specified by the operator using the keyboard 48. Read and copy (record) to floppy disk 43. The graph creation means 47 creates image data for displaying the measurement data designated by the operator as a graph.
[0056]
The image synthesizing means 46 synthesizes the original image, the created trend graph of the measurement data, and the current measurement data (interface position), and displays them on the TV monitor 5. In the image measurement system shown in FIG. 2, a trend graph is automatically created and displayed on the TV monitor 5, and all measurement data is automatically saved in the hard disk 42 or floppy disk 43, and further specified by the operator. The date measurement data can be immediately displayed on the TV monitor 5. Thereby, an operator's burden can be reduced significantly.
[0057]
Next, the peripheral processing means 45 has a self-diagnosis function when there is an abnormality, which will be described below. I tried to acquire an image with the TV camera 3, but if the image signal (synchronization signal) does not come in continuously for the set number of times or more, I suspect that the TV camera 3 is faulty or the cable is broken. A warning message is displayed on the CRT monitor 6. The reason why the number of times is equal to or more than the set number is that when the cable is long and is input to the image analysis means 40 via the outdoors, the signal may be instantaneously disturbed by external noise or the like. Even if the measurement data cannot be obtained because the image cannot be acquired well due to such a momentary abnormality, the previous measurement data can be retained in that case, and the subsequent measurement can be continued. . In this way, a warning display due to a momentary abnormality is not performed.
[0058]
Further, it is possible to self-diagnose a failure of the image input board. This performs self-diagnosis of the A / D conversion chip of the A / D conversion unit 10 (see FIG. 3). For example, if the image acquisition is complete, but the actual image data is completely flat image data (all pixels are 0 values or all pixels are 255 values), it is a chip failure. Therefore, a warning is displayed on the CRT monitor 6 also in this case.
[0059]
Self-diagnosis can also be performed based on statistical data. The above-mentioned failure of the TV camera 3 or A / D conversion chip is a warning display only when the set number of times continues, but it is not normal. If it cannot be ignored, a warning is displayed on the CRT monitor 6 as well.
[0060]
Next, a display screen of the CRT monitor 6 is shown in FIG. In the present embodiment, a TV monitor 5 and a CRT monitor 6 are provided as display means, but the TV monitor 5 displays a real-time original image taken by the TV camera 3, an interface position, a trend graph, and a date display. This is usually sufficient for information.
[0061]
The CRT monitor 6 is used when operating the computer of the image analysis means 4. The trend graph C2 and date time C5 in the CRT monitor 6 are the same as the display screen of FIG. The status window C1 displays an operation state of the image measurement system, a message for an operator's key operation, and a warning display in case of abnormality. In FIG. 16, “interface measurement: normal operation” is displayed. In addition, when the measurement data is copied to the floppy disk 43, “data retrieval” is displayed, and “TV signal abnormality message” and “image input board abnormality message” when an abnormality occurs in the system are displayed.
[0062]
The analysis result display window C3 displays a breakdown of image processing results for each time. The breakdown is updated every predetermined time. The operator usually does not need to pay attention to the display contents of the window C3, and is mainly used for diagnosis at the time of abnormality. In this window C3, the numerical value indicates the number of occurrences.
The disk usage status display C4 displays the current usage status of the hard disk 42. The measurement data is saved in the hard disk 42 every day in a predetermined file format. However, when the month changes, the measurement data for each month is collected into one separate file. In addition, measurement data from the first day of this month to today can be taken out by a predetermined key operation.
[0063]
<Modification>
In the present embodiment, the extracted image is a ternary image. However, in the case of a simple interface as shown in FIGS. 11 and 12, it may be extracted as a binary image.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration including an interface measuring device.
FIG. 2 is a block diagram showing the configuration of an image analysis system
FIG. 3 is a block diagram illustrating functions of image analysis means.
FIG. 4 Operation flowchart
FIG. 5 shows the current original image
FIG. 6 shows a past original image.
FIG. 7 is a diagram showing a difference image
FIG. 8 is a density histogram of a difference image.
FIG. 9 shows an extracted image
FIG. 10 is a diagram showing a density distribution pattern.
FIG. 11 is a model 1 of a density distribution pattern.
FIG. 12 is a model example 2 of a density distribution pattern.
FIG. 13 is a model example 3 of a density distribution pattern.
FIG. 14 is a model 4 of a density distribution pattern.
FIG. 15 shows a display screen example of a TV monitor.
FIG. 16: Display screen example of CRT monitor
FIG. 17: Example of input data
FIG. 18: Example of teacher data
[Explanation of symbols]
3 TV camera (image acquisition means)
4 Image analysis system
5 TV monitor
6 CRT monitor
40 Image analysis means
40b Difference calculation unit
40c Smoothing processor
40d threshold value calculation unit
40e Change extractor
40f Change extraction image evaluation unit
40g Interface position determination unit

Claims (8)

In an interface measurement apparatus comprising an image acquisition means for acquiring an original image including an interface to be measured and an image analysis means for analyzing the acquired original image to obtain an interface position, the image analysis means is different. A difference calculation unit that obtains a difference image between original images acquired at time, a change extraction unit that obtains an extracted image obtained by extracting a change of the original image based on the difference image, and an interface position based on the extracted image An interface position determination unit for determining, and a change extraction image evaluation unit that determines whether the change extraction image is due to a change in the interface with respect to the change extraction image obtained in the interface measurement process; An interface measuring device characterized by comprising. Prior to obtaining the difference image between the original images acquired at different times, the image analysis means analyzes the positional deviation between the current original image and the past original image, and corrects the deviation in the direction of correcting the deviation. The interface measuring apparatus according to claim 1, further comprising a position deviation correction unit that obtains a differential image after translating at least one of them. The interface measurement apparatus according to claim 1, wherein the image analysis unit further includes a smoothing processing unit that obtains a smoothed image obtained by removing a noise component from the difference image. The threshold value calculating part which calculates | requires the threshold value for calculating | requiring the said extracted image from the density | concentration histogram of the said difference image or the density | concentration histogram of the said smoothed image is further provided. The interface measuring device according to claim 1. The change extraction unit obtains the extracted image from the difference image or the smoothed image using a threshold value obtained by the threshold value calculation unit or a preset threshold value. Item 5. The interface measurement apparatus according to any one of Items 1 to 4. The interface position determination unit determines an interface position by comparing a density change pattern obtained from the difference image with a density change pattern prepared in advance. The interface measuring device according to claim 1. Display means for displaying the original image acquired by the image acquisition means, the display means displaying the interface position determined by the interface position determination unit and a graph showing the change in the interface position over time together with the original image. The interface measuring device according to claim 1, wherein the interface measuring device can be used. The image analysis means prepares a plurality of past original images for obtaining a difference image with respect to the original image to be subjected to interface measurement, performs image analysis on each past original image, The interface measurement apparatus according to claim 1, wherein the interface position is determined by adopting a most preferable image analysis result from among the image analysis results.

Images