JPH08123964A - Method and device for registering model pattern - Google Patents

Abstract

PURPOSE: To provide the model pattern registration device which can register a model pattern based upon an image in a frame by securely setting a specific area in an image displayed on a display device as the frame for registration through simple processing. CONSTITUTION: An image of a field at a monitor destination which is picked up by a video camera 10 is displayed on a monitor TV 44 through a video image input part 20 and an image processing part 30. A touch panel 41a is mounted on the monitor TV and an operator touches the center of an object to be registered (e.g. trespasser) on the screen. The coordinates (reference coordinate) of the contact point is detected by the touch panel and transferred to a CPU 32. The CPU 32 extracts the edge of the object to be registered, makes a search from the reference point in an X and a Y direction and detects the edge part, and decided with the detected lines as frame sides, thereby automatically setting the frame; and then the model pattern is generated on the basis of the image data in the frame through normal processing and registered.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a model pattern registration method and device.

[0002]

2. Description of the Related Art Generally, when a predetermined object or person existing in an inspection (monitoring) area is detected by pattern matching, a model pattern serving as a comparison reference is registered in advance, and the model pattern is determined. The degree of similarity and the degree of conformity with the target image is obtained, and it is determined whether the target object.

In order to register such a model pattern, the picked-up image data is displayed on the monitor, and in that state, a frame surrounding a predetermined portion of the image data is set. Since this frame normally has a rectangular shape, it is performed by designating a start point and an end point (a pair of vertices on a diagonal line) on the display screen.

[0004]

However, in the conventional frame setting method, the pointer (cursor) is moved using a mouse or the like to specify the coordinates of two points separated by a predetermined distance. It requires three steps of instruction at least twice and movement processing between the two points, the frame cannot be set instantaneously, and as a result, it takes time to register the model image.

For example, when a predetermined portion of an image displayed on a monitor which is picked up by a video camera and transmitted online and registered as a model pattern, the object to be processed is moving, a frame Since the object being set moved, the frame could not be set correctly, and as a result, it was difficult to register the model pattern online.

Further, if a large frame is set with a margin, useless background parts are introduced, the memory capacity for registering as a model pattern becomes large, and the subsequent matching process also takes time, resulting in poor accuracy.

Further, if the size of the frame to be registered is set as a fixed value in advance, it is not possible to correctly cope with the case where the size of the object actually imaged and displayed changes. Model pattern cannot be generated).

Further, it is better that the frame to be set is circumscribing the object, but since the vertex of the diagonal line is apart from the edge of the object, it is difficult to set the frame so as to make accurate contact.

The present invention has been made in view of the above background. An object of the present invention is to solve the above problems and to create a model pattern for a predetermined region in an image displayed on a display device. When registering, you can set the frame to be registered with a simple process in a short time, and even if the object is moving, you can reliably cut out the frame and display the image in the frame after that. It is an object of the present invention to provide a model pattern registration method and device capable of registering a model pattern based on the model pattern registration method.

[0010]

In order to achieve the above-mentioned object, in a model pattern registration method according to the present invention, in the method of registering a predetermined area in an image displayed on a display device as a model pattern, the image A predetermined position in the inside is designated and received as a reference coordinate, the difference between the image to be processed and another image where the registration target does not exist or is at a different position is calculated, and the difference image is obtained. Then, a search is performed in a predetermined direction from the given reference coordinates to detect a boundary to be registered, a predetermined frame is set based on the detected boundary, and a model is created based on image data existing in the set frame. The pattern is generated and registered.

[0011] Preferably, the reference coordinates specify substantially the center position of the registration target, and a search in a predetermined direction from the reference coordinates is performed in the vertical and horizontal directions from the reference coordinates, and one boundary is If not detected and the boundary in the opposite direction is detected, it is assumed that the undetected side has a boundary at a position approximately equal to the distance from the reference position to the detected boundary in the opposite direction. However, the detected or assumed boundary is set as a frame side forming the frame.

The boundary is, for example, an edge portion extracted by performing a predetermined filtering process on the difference image, and the boundary may be detected based on the edge strength. Alternatively, a portion in which the density or the density difference of the grayscale images forming the difference image is equal to or less than a predetermined value may be determined as the boundary.

Further, after setting a frame by any of the above methods, the frame is displayed at a predetermined position on the display device so as to be superimposed on the image, and the coordinate input of a predetermined point in the image is received again. In this case, if the newly input coordinates are the same as the reference coordinates, correction is performed to change the size of the set frame, and / or the coordinates of a predetermined point in the image are input again. If the newly input coordinates are different from the reference coordinates, it is preferable to perform the correction of moving the set frame based on the newly input coordinates.

In the model pattern registration apparatus according to the present invention suitable for carrying out the above registration method, a display device for displaying a predetermined image and coordinate input means for inputting coordinates of an arbitrary point on the display device. An edge extraction means for extracting an edge to be registered, a search is performed in a predetermined direction from the coordinates given by the coordinate input means, the extracted edge portion in the search direction is detected, and based on the edge portion Frame setting means for setting a frame for pattern registration, and means for generating a model pattern based on the image data in the frame set by the frame setting means and registering it in the storage means.

Further, a display device for displaying an image including a registration target, coordinate input means for inputting coordinates of an arbitrary point on the display device, the image to be processed, and the registration target image do not exist or are different. A difference image generation means for obtaining a difference from a predetermined comparison image existing at a position, and a search is performed in a predetermined direction from coordinates given by the coordinate input means based on the difference image generated by the difference image generation means, A frame setting unit that detects the boundary of the object based on the density in the search direction or the difference in density before and after, and sets a frame for pattern registration based on the boundary, and a frame set by the frame setting unit. A model pattern may be generated on the basis of the image data in the above and registered in the storage means.

The coordinate input means is preferably a touch panel mounted on the display device. When the frame setting means further includes a correction function for correcting at least one of the size and the position of the frame set based on the correction command signal given from the correction command input means. , And more preferable.

[0017]

When a registration target as a model pattern exists in the image output on the display device, predetermined coordinates (reference coordinates) in the registration target are designated using the coordinate input device. The predetermined coordinates are preferably the central portion of the pattern to be registered. Further, when a touch panel is used as the coordinate input device, the reference coordinates can be designated easily, quickly and accurately by touching a screen on the display device, that is, a predetermined portion of the registration target displayed.

Given the reference coordinates, the difference between the image to be processed and a predetermined image to be compared is calculated. Then, the value of the pixel forming the background portion with or without movement becomes small, and the registration target (the boundary portion) is extracted. Then, when predetermined filtering is performed, the edge portion is emphasized. Therefore, a search is performed from the reference coordinates in a predetermined direction, for example, in the vertical and horizontal directions, and the edge portion to be registered is extracted. Then, if extracted, a frame for model pattern registration is set based on the edge portion.

Specifically, for example, when the edge portion is extracted, the edge portion is set as the side of the frame. As a result, a frame for automatically suppressing the existence of unnecessary image portions such as the background portion is automatically set with the minimum necessary registration target. In addition, depending on the shape of the registration target and the imaging state at that time (relative positional relationship with other objects, etc.), the edge may not be extracted. In that case, a portion where an edge virtually exists is set based on another detected edge portion, and that portion is set as a side of the frame. As a result, even if an edge is not detected, the remaining frame sides are automatically set based on the other parts where the edge is detected, and a closed space configured by connecting each frame side is registered. It becomes a frame for. After that, the image data existing inside the frame is subjected to predetermined image processing (the image as it is or a predetermined feature amount is extracted) to generate / register a model pattern.

Even if such filtering is not performed, since the density of the background portion is small, it is possible to set the frame by performing a search in a predetermined direction and detecting the density of each pixel. That is, as a result of the search, it can be assumed that there is no density difference (small) or the density becomes 0 or small as the background. Therefore, such a part may be the boundary of the registration target, and a predetermined frame may be set. (Other than setting the frame,
It is carried out by the same operation as above). As a result, although the accuracy is reduced, the filtering process is not required, so that the configuration is simple and the process can be performed at higher speed.

Furthermore, when a predetermined correction function is added, the correction (size and existing position) for the frame determined by the above operation can be changed, and a more desired frame can be set,
The pattern registration based on it is also performed accurately.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of a model pattern registration method and apparatus according to the present invention will be described in detail below with reference to the accompanying drawings. FIG. 1 shows an overall configuration diagram of an embodiment of a model pattern registration device in which the present invention is implemented. Then, in this example, an example applied to the case of registering a person or other object that has entered the site (monitoring area) such as a surveillance camera as a model pattern.

As shown in the figure, the output (video signal) of the video camera 10 which is an image pickup device for picking up the surveillance area is connected to the video image capturing section 20, and predetermined image data based on the actual image is digitized. And stored.
Then, the predetermined image data (digital) captured by the video image capturing section 20 is transferred to the image processing section 30, and predetermined processing is executed there according to the coordinate instruction or the like given from the input devices 41 and 41a. Then, a frame indicating an area for automatically registering the model pattern is set, and the model pattern is generated based on the image data in the frame.

The generated model pattern is registered in the main memory 31 built in the image processing unit 30 or a storage medium in an external storage device such as the HDD 42 or the FDD 43 connected to the image processing unit 30 and required. It is read out in accordance with the above, is displayed at a predetermined position on the monitor TV 44, and is used when matching with the image data of the monitoring area displayed on the monitor TV 44.

Next, each part will be described in detail. In the video image capturing part 20, the video signal (analog) sent from the video camera 10 is received by the video input part 21, separated into horizontal / vertical synchronizing signals, sampled and then A / D
The brightness level is digitized by the converter 22, and X,
It is sequentially stored in the video memory 23 according to the Y address. Then, in the video memory 23, the video camera 1
The image data of the site (monitoring area) obtained by digitizing the video image captured at 0 is stored.

As the monitor TV 44, a color liquid crystal display is used in this example, and a touch panel 41a as a reference point input means is mounted on the surface. It should be noted that the present invention is not limited to the liquid crystal display described above, and other display devices such as a CRT may of course be used.

The input device 41 can use various devices such as a keyboard, a mouse, a trackball, and the like.
Various commands required when registering a model pattern can be input.

In the image processing section 30, the photographed actual image data (digital image) stored in the video memory 23 is transferred to the graphic memory (frame memory) 33 in a through state as it is in accordance with a control command from the CPU 32. , The data temporarily stored in the memory 33 is D
A / A converter 34 converts to analog signal, then monitor TV
44 is displayed. In addition, each of the external devices 41 to 43 described above
Is an I / O control 3 built in the image processing unit 30.
5 and is configured to transmit and receive information via the I / O control 35. Further, the coordinate signal of the touched portion from the touch panel 41a is similarly sent to the CPU 32 via the I / O control 35.

The function of the CPU 32, which is the main part of the present invention, will now be described with reference to the flowcharts of FIG. Normally, the image data (current image) of the surveillance area picked up and sent by the video camera 10 is directly sent to the monitor T.
Output to V44 (ST1). Then, when registering the model pattern, the operator specifies the coordinates of the central portion of the target object to be registered. Specifically, in this example, since a touch panel is mounted on the display screen of the monitor TV 44, it is performed by touching a predetermined position on the touch panel 41a. Therefore, the CPU 32 uses the touch panel 4
It is determined whether or not the operator's finger or the like touches 1a, and when touched (pressed), the indicated X and Y coordinates are displayed while the display screen at that time is temporarily stopped (frozen). Acquire (ST2-4). In addition, the suspension is
By not updating the contents of the image data stored in the graphic memory 33, the same image can be continuously output.

Next, the process proceeds to a frame (window) setting process for registering the model pattern. First, a predetermined preprocessing is performed on the displayed image to extract the edge portion of the registration target pattern (ST5). The specific processing of this step 5 is as shown in FIG.
That is, this processing step 5 (flow shown in FIG. 3)
Constitute the edge extraction means in the present invention.

Specifically, it is first determined whether or not the object to be registered is moving (ST10). This judgment is made based on a manual input from the input device 41 (instruction by the operator). The case where the intruder is moving means, for example, a phenomenon in which the intruder is moving in the surveillance area, and the case where the intruder is not moving is such that the intruder is stopped at a certain point in the surveillance area. Means a phenomenon.

When moving, the CPU 32 takes in the image data (frame) stored in the graphic memory 33 to be processed and the image data transferred next from the video memory 23 (continuously. To do 2
(Take in one frame) and obtain the difference between the two images (ST11). On the other hand, when it is stationary, the difference between the reference image (the image of the normal monitoring area where there is no intruder, etc.) previously captured and held and the image data to be processed is obtained ( ST12). As a result, the background portion that does not change (move) is eliminated.

However, if left as it is, there is noise such as an isolated point, and the edge portion of the object is blurred. Particularly when the difference between two consecutive frames when the object is moving is calculated, the edge portion of the object has a predetermined width according to the moving distance. Therefore, a predetermined filtering is performed to remove noise such as the isolated point,
The edge is sharpened (ST13). Note that the processing is performed by applying a Laplacian filter or other predetermined filter to the difference image data, but the specific coefficients of the filter can be basically conventional ones. Also, the edge portion can be extracted with higher accuracy by appropriately modifying the filter coefficient, the number of squares, and the like as necessary.

In this example, when the boundary portion of the object to be registered is detected, post-difference filtering is performed to cause a clear difference in edge strength. However, the present invention is not limited to this, and the processing target is not limited to this. Since the image is taken as a predetermined image (between two frames or a reference image), the background portion can be removed with relatively high accuracy (the brightness becomes 0 or a very small value) only by the difference processing. ,
The grayscale image data is searched for a grayscale change without filtering, and the change disappears (becomes less than or equal to a predetermined threshold value), or the brightness is "0" or less than or equal to a predetermined threshold value. It is also possible to determine that the broken portion is the boundary of the object.

Next, returning to step 6 shown in FIG. 2, the center (origin) is set to the reference point coordinate (input coordinate) which is the reference for frame setting.
Assuming an X-Y axis, the edge position of the object existing on the axis is detected, and the frame is set based on the detected position. Then, the specific process of step 6 is as shown in FIG. That is, this processing step 6
(The flow shown in FIG. 4) constitutes the frame setting means in the present invention.

First, the vertical direction (Y
(Axis), and the existence position of the edge on the Y axis is detected (ST21). In this search, in this example, a window having a width of a predetermined unit (10 pixels) in the search direction, that is, a direction perpendicular to the Y axis (the search direction is 1 pixel and the center of the window is the Y axis) is set, and the search direction is set. The state of the line is searched by moving one pixel at a time to determine whether or not there is a change, and the line immediately before when there is no change is the position where the edge exists. That is, in the pre-processing in step 5, the value of the pixel corresponding to the background portion which is at least unchanged or small is small. Therefore, scratches in the window are binarized with a predetermined threshold value, and the number of pixels having a value exceeding the threshold value existing in the window is counted for each search. Then, it can be assumed that the background portion has been reached with little or no change. Then, in order to prevent unnecessary image data such as the background portion from existing in the frame, the line on the immediately preceding (input coordinate (center) side) is edged as described above in order to set the frame as small as necessary. And

If the edge portions are extracted on both the upper and lower sides, the line portions are set as the upper limit (upper side) and the lower limit (lower side) of the frame (ST22, 23). On the other hand, if an edge portion is detected on only one side, the detected side is set as the frame side that constitutes the frame and the frame side that constitutes the undetected side is set based on the detected line. To do. In this example, the point-symmetrical position with respect to the input coordinate, that is, the position where the edge is not detected is a position separated by the same distance as the distance from the input coordinate to one edge portion detected. To the frame edge of (ST2
4, 25). If the target is an intruder (human),
Generally, the model pattern to be registered is a human face, so
The upper limit edge is detected because it is above the head (vertex), but since the body exists on the lower side, it is difficult to detect the edge portion. Therefore, the normal flow is to jump from step 22 to step 24, but in the process of step 25,
The lower limit will be set based on the detected upper limit.

Next, from the input coordinates to the left-right direction (X-axis direction)
Is searched to extract the edge part (ST2
6). The specific processing of this search is performed in the same manner as in the vertical search in step 21 described above (searching for 10 pixels each from the input coordinates), except that the direction is different.

Then, similarly to the processing after the search in the vertical direction, if both edges are extracted, each line is set as the left limit (left side) and the right limit (right side) of the frame (S).
T28). If only one side is detected, the detected side is the detected part, and the non-detected side is the part equidistant from the edge part detected based on the input coordinates on the opposite side. Frame sides (ST29, 30).

In addition, for example, an intruder to be processed may overlap another person or the shadow of an installation object, and both the left and right edge portions may not be detected. In this case, the left and right frame sides are determined based on the upper and lower frame sides determined by the processing of steps 21 to 25 (ST31). That is,
When the object is a human, the most efficient model pattern can be generated by setting a vertically long rectangular frame. Therefore, based on a predetermined ratio (aspect ratio, set via the input device 41), Calculate the width for the vertical length,
A point on the X-axis having the width and having the input coordinates as the center is obtained (a point separated by half the width), and the portion is set on each of the left and right frame sides.

As a result, since the upper, lower, left and right frame sides are set, a rectangular frame body centering on the input coordinates can be automatically set, and the set frame can be used as an actual image stored in the graphic memory 33. The image data is displayed so as to be superimposed on the base image data (ST32). At this time, the place touched by the finger as a reference point is indicated by a "+" mark.

When the determination in step 24 is No, that is, when the upper and lower edge portions are not detected, an error (frame setting is impossible) is determined in this example, and the process is terminated (ST33). In this example, intruders (especially humans) who enter the monitoring area are registered in real time as a model pattern, and then tracking processing is automatically performed, and the presence or absence of the same person in the monitoring area is detected. Therefore, since it is applied when registering a model pattern online, a human must be able to detect the apex of the head without fail. Therefore, when the answer is No in step 24 as described above, the head part is hidden or is incorrect. In addition, since it is highly likely that an inappropriate portion was designated as the input coordinate, error processing was performed.

However, when the objects to be processed are different, at least one side is detected regardless of whether the edge part in the horizontal direction is first detected or the search order in the vertical and horizontal directions is not performed without making the above determination. If the edge part of is detected,
The non-detected side may be automatically set based on the detected position of the predetermined side or the like (for example, using the processing of steps 25, 30, 32).

If the frame is automatically set in this way, the process returns to step 7 shown in FIG. 2 to judge the presence / absence of a correction command. If there is no correction command, the frame automatically set is officially set. A model pattern registration frame, and the image data existing in the frame (image data based on an actual image stored in the graphic memory 33 in a paused state) is extracted as it is or a predetermined feature amount is extracted as a model pattern. Convert and register (ST7, 8). Since the conventional method can be applied to the registration processing based on the image data in the frame, detailed description thereof will be omitted.

On the other hand, in this example, the operator who sees the automatically set frame (displayed at the predetermined position in step 32) judges the suitability and corrects the position and size of the frame semi-manually. I am able to do that. Specifically, when any part of the displayed screen is touched, the process proceeds to the correction process (ST9).

The specific processing flow of the correction processing is as shown in FIG. That is, first, it is judged whether or not there is a coordinate input from the touch panel 41a within a fixed time (ST40), and if not touched, the size and position of the frame at that time are determined, and the process returns to step 8.

When there is a coordinate input, it is determined whether or not the input position is the same as the input coordinate (which is indicated by the "+" mark at the center of the current frame) pressed first (or last time). Judgment (ST41), if it is the same place, it is judged to be a frame correction command (ST42), and X, Y directions (vertical and horizontal directions).
Then, the frame is expanded by the reference pixel (10 pixels in this example) (ST43). As a result, the existing positions of the frame sides are expanded by 10 pixels while keeping the center position the same. Therefore, the vertical direction and the horizontal direction are expanded by 20 pixels each.

Then, until the separation distance of the frame side from the input coordinate in the X direction or the Y direction spreads to MAX (for example, 150 pixels), the frame is expanded every time the input from the touch panel is made (ST43, 44). When it spreads over a certain value, it is judged that the request from the operator is desired to be smaller than the automatically set frame, and the X and Y directions (both positive and negative directions) are each a predetermined number of pixels (for example, 100 pixels). After reducing the size (ST45), the process returns to step 40. After that, if the same place is continuously touched, the processes of steps 41 to 44 are repeated, so that the pixels spread in the vertical and horizontal directions by 10 pixels from the MIN frame set in step 45.

That is, for example, the automatically set frame is shown in FIG.
If it is shown by a solid line, the frame gradually becomes large as shown by a broken line in the figure, such as →→→ ... On the other hand, if it is touched even if it spreads to MAX, MI is shown as indicated by the chain double-dashed line in the figure.
It becomes the frame of N. After that, by repeating this process, the frame is expanded from the MIN position by a predetermined number.

On the other hand, if a place different from the previous input place is touched, the step 4 after the branch judgment of the step 41
Jumping to step 6, the frame is moved so that it is determined to be a frame moving command and the newly given coordinate point becomes the center (ST46, 4).
7) After that, it returns to step 40. The size of the frame is not changed during this movement.

That is, for example, if an automatically set frame (center coordinates (X0, Y0) is as shown by the solid line in FIG. 7, the next input coordinate (X,
If Y) is deviated, the position of the frame moves as shown by the broken line in the figure.

If there is no coordinate input from the touch panel for a certain period of time, it is determined that the frame is a formal model pattern registration frame based on the size and position of the frame displayed at that time.
Returning to step 8, the model pattern registration process is performed. In addition, based on the compulsory registration signal from the input device 41 or the touch panel 41a, registration may be performed in the state of the frame at that time without waiting for a certain period of time.

In the above embodiment, the input coordinates (reference position) is instructed by the touch panel 41a.
The present invention is not limited to this. For example, based on an instruction from a mouse, a trackball, a keyboard, or the like, a cursor or a pointer is moved on the screen, and a confirmation instruction (click, return, etc.) is sent when a predetermined position is reached. May be. However, when the touch panel is used, it is not necessary to move the cursor and the like, and the instruction input can be performed in a short time, and since the displayed position is directly touched, the input instruction of the desired position can be performed easily and accurately.

Next, one embodiment of the model pattern registration method according to the present invention will be described using the above-mentioned embodiment. Image data picked up by the video camera 10 and transferred is displayed on the monitor TV. Then, in a normal state, it is assumed that the scenery as shown in FIG. 8 is displayed on the monitor TV 43. The CPU 32 stores the image data in this state in an internal buffer or a predetermined storage means.

During monitoring in this state, the intruder M1 is displayed on the monitor TV 44 at a certain timing, for example, as shown in FIG. 9, and the face of the intruder M1 is registered as a model pattern. The monitor T
Touch the center of the face on the screen of V44. Then, Figure 2
According to the flowchart shown in FIG. 4, the frame W centering on the input coordinate point is obtained and displayed on the monitor TV 43 together with the touched position (input coordinate point “+” mark).

If the operator can leave the displayed frame W as it is, the image in the frame W is registered as a model pattern. Further, when it is desired to correct the display position, a predetermined position (the same position as “+” or a different position) is touched. Then, according to the flowchart shown in FIG. 5, the size and position of the frame are corrected.

In each of the above-described embodiments, the face of the intruder in the area is registered as a model pattern based on the image of the area to be monitored sent from the video camera (monitor camera), and a predetermined pattern thereafter is registered. Processing (By performing tracking processing while matching and detecting the same person,
An example of using it as a model pattern that serves as a reference for comparison when detecting the existence time in the area and determining whether or not it is an abnormal situation has been described, but the present invention is not limited to this. , And can be used for creating / registering model patterns when performing various types of matching. Moreover,
Although the present embodiment has a feature (effect) that online registration can be performed for an image being picked up by a video camera, the present invention is not limited to this, and an arbitrary method (digital still camera or image It can also be applied to the case of registering a predetermined part in the image acquired and generated by a device such as a scanner that captures an image of a stationary object and displayed on the display device as a model pattern. .

[0058]

As described above, in the model pattern registration method and device according to the present invention, one point (for example, the center point) of a portion to be registered as a model pattern is set as a reference coordinate in the image displayed on the display device. Just tell
It is possible to automatically set a predetermined frame including the registration target, generate a model pattern from the image data existing in the set frame by predetermined image processing, and register the model pattern. That is, the operator who performs the registration process can automatically register the desired model pattern only by performing the process of designating the reference coordinates. Since the computer automatically processes the frame, the setting of the frame is also speeded up. Therefore, for example, a model pattern is registered online with respect to a predetermined portion (registration target) in a screen which is imaged by an imaging device and transmitted online and displayed on a display device, regardless of whether or not the object is moved. be able to.

When a touch panel is used as coordinate input means for designating the reference coordinates (claim 9).
The reference coordinates can be input more accurately and more easily, which improves the processing speed and the accuracy of the processing result.

Furthermore, in the case where a predetermined correction process is performed on the automatically set frame (claims 5, 6, 1).
In 0), since the frame that specifies the actual registration range and the image to be processed are displayed in a superimposed state, it is possible to easily determine the suitability of the automatically set frame, and determine the existing position and size of the frame. Can be easily corrected to the optimum position.

[Brief description of drawings]

FIG. 1 is a block diagram showing a preferred embodiment of a registration device according to the present invention.

FIG. 2 is a flowchart showing functions of a CPU used in this embodiment.

FIG. 3 is a flowchart showing functions of a CPU used in this embodiment.

FIG. 4 is a flowchart showing functions of a CPU used in this embodiment.

FIG. 5 is a flowchart showing functions of a CPU used in this embodiment.

FIG. 6 is a diagram illustrating an operation of a method for correcting a frame size.

FIG. 7 is a diagram illustrating the operation of the method for correcting the position of the frame.

FIG. 8 is a diagram showing a display screen of the monitor TV displayed in each step of the embodiment of the model pattern registration method.

FIG. 9 is a diagram showing a display screen of a monitor TV displayed in each step of the embodiment of the model pattern registration method.

FIG. 10 is a diagram showing a display screen of a monitor TV displayed in each step of the embodiment of the model pattern registration method.

[Explanation of symbols]

 10 Video Camera 20 Video Image Capture Unit 30 Image Processing Unit 31 Main Memory (Storage Means) 32, 32 'CPU (Edge Extraction Means, Frame Setting Means) 33 Graphic Memory 34 D / A Converter 41 Input Device 41 Touch Panel (Coordinate Input Means) 42 HDD (storage means) 43 FDD (storage means) 44 Monitor TV (display means)

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location 9061-5H 15/70 465 Z

Claims (10)

[Claims] 1. A method for registering a predetermined area in an image displayed on a display device as a model pattern, wherein a predetermined position in the image is designated and received as reference coordinates, and the image to be processed and the registration are performed. The difference between the target image and another image at a different position is determined, and the boundary of the registration target is detected by performing a search in the predetermined direction from the given reference coordinates in the obtained difference image, A model pattern registration method in which a predetermined frame is set based on the detected boundary, and a model pattern is generated and registered based on image data existing in the set frame. 2. The reference coordinates specify substantially the center position of the registration target, and a search in a predetermined direction from the reference coordinates is performed in the vertical and horizontal directions from the reference coordinates, and one boundary is detected. However, when a boundary in the opposite direction is detected, it is assumed that the non-detected side has a boundary at a position substantially equal to the distance from the reference position to the detected boundary in the opposite direction, A model pattern registration method in which the detected or assumed boundary is set as a frame side forming the frame. 3. The boundary is an edge portion extracted by performing a predetermined filtering process on the difference image, and the boundary is detected based on the edge strength. The model pattern registration method described in. 4. The model pattern registration method according to claim 1 or 2, wherein the grayscale image forming the difference image is determined as the boundary at a portion where the density or the density difference is equal to or less than a predetermined value. . 5. After setting a frame by the method according to any one of claims 1 to 4, the frame is displayed at a predetermined position on the display device so as to be superimposed on the image, and the coordinates of a predetermined point in the image are displayed again. 5. When receiving an input, if the newly input coordinate is at the same position as the reference coordinate, correction is performed to change the size of the set frame. The model pattern registration method described in the section. 6. After setting a frame by the method according to any one of claims 1 to 5, the frame is displayed at a predetermined position on the display device so as to be superimposed on the image, and the coordinates of a predetermined point in the image are displayed again. 3. When receiving an input, if the newly input coordinates are different from the reference coordinates, correction is performed to move the set frame based on the newly input coordinates. 5. The model pattern registration method according to any one of items 5 to 5. 7. A display device for displaying an image including a registration target, coordinate input means for inputting coordinates of an arbitrary point on the display device, edge extraction means for extracting an edge to be registered, and the coordinate input. A frame setting means for performing a search in a predetermined direction from coordinates given by the means, detecting the extracted edge portion in the search direction, and setting a frame for pattern registration based on the edge portion; A model pattern registration device comprising means for generating a model pattern on the basis of image data in a frame set by the setting means and for registering it in a storage means. 8. A display device for displaying an image including a registration target, coordinate input means for inputting coordinates of an arbitrary point on the display device, the image to be processed, and the registration target image not present or different from each other. A difference image generation unit that obtains a difference from a predetermined comparison image existing at a position, and a search is performed in a predetermined direction from the coordinates given by the coordinate input unit based on the difference image generated by the difference image generation unit. A frame setting unit that detects the boundary of the object based on the density in the search direction or the density difference before and after, and sets a frame for pattern registration based on the boundary, and a frame set by the frame setting unit. A model pattern registration device having means for generating a model pattern on the basis of image data in the inside and for registering it in a storage means. 9. The model pattern registration device according to claim 7, wherein the coordinate input means is a touch panel mounted on the display device. 10. A correction function further comprising a correction command input means, wherein the frame setting means corrects at least one of a size and a position of the frame set based on a correction command signal given from the correction command input means. The model pattern registration device according to claim 5, wherein the model pattern registration device includes: