JP2006172381A - Image observation method and device for performing the same method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method and a device of observing an image which divides an object area of a color distance image into binary by a single color distance threshold. <P>SOLUTION: In this image observation method and the device, when an observation object area is extracted by dividing a surface image of an observation object into object areas with similar hue, structure for converting RGB signals of the surface image obtained by picking up the surface image into a numeric value called as a color distance to convert the surface image into the color distance image and dividing the object area of the color distance image into binary by a single set color distance threshold is adopted, though three thresholds are required in a conventional method for decomposing the RGB signals into a luminance signal and a chromaticity signal, performing the image processing of the luminance signal to divide the object area, a single color distance threshold suffices for a threshold, the color distance threshold creates color distance histogram in each object area frame and the color distance threshold suitable for the extraction and division processing of the object area is properly set using an inflection point of the histogram as a guidepost. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image observation method and an apparatus for performing the method, and in particular, when extracting an observation target area by dividing a surface image of an observation target into target areas having similar hues, the image is captured by a CCD color camera. The RGB signal of the surface image obtained is converted into a numerical value called color distance to convert the surface image into a color distance image, and the target area of this color distance image is binarized with a set single color distance threshold. The present invention relates to an image observation method that employs a segmenting configuration and an apparatus that implements the method.

  In general, it is rare that the observation target area of a graphic pattern to be extracted in image observation can be fixed to a specific color or a specific shape, and usually a combination of target areas of multiple colors and multiple shapes. It is predicted that The target graphic pattern can be recognized by extracting the plurality of target areas and combining the plurality of target areas. If the number of target areas is large, the observation target area can be approximated more accurately to the target graphic pattern. In general, a technique for automatically finding a human face in an image that has been captured by a camera and electronically binarized, or checking the characteristics of the face to confirm the presence of the face is disclosed in Patent Document 1. ing.

When dividing a color image obtained by imaging the surface of an observation target into target areas having similar hues, three types of filters R (red), G (green), and B (blue) are placed in front of the objective lens of the CCD camera. By arranging and switching these, for R (red), G (green), and B (blue), an 8-bit R luminance signal, an 8-bit G luminance signal, and an 8-bit B luminance signal are combined. In addition, a 24-bit signal of a three-dimensional vector called an RGB signal is output from the CCD color camera.
The 24-bit RGB signal output of the CCD color camera is then divided into R (red), G (green), and B (blue) colors and processed into three monochrome images.
JP 2002-203239 A

  By the way, as a method of dividing a color image into target regions, there is a method in which a 24-bit RGB signal is decomposed into luminance signals and chromaticity signals called YCrCb signals, and the luminance signals are subjected to image processing and divided into regions. However, in this method, the area is substantially divided as a monochrome image, and the color information is not used and is deleted. That is, although the original image is a color image, the area division of the color image is performed by converting to a monochrome image and setting a binarization threshold value to determine whether it is white or black. This is because, in this method, when a color image is divided into regions, three threshold values are required as threshold values, and it is impossible to set each threshold value to have the meaning of luminance, hue, and saturation.

  The present invention converts RGB signals into numerical values called color distances to convert a color image into a color distance image, and binarizes the color distance image by setting a single appropriate color distance threshold. The present invention provides a method for observing an image and a device for implementing the method, in which a configuration for dividing a target area is adopted to solve the above-described problems.

Claim 1: The RGB signal of a surface image obtained by imaging with a CCD color camera is converted into a color distance, and the target area is binarized with a single color distance threshold value, so that the hue of the surface image of the observation target is obtained. An image observation method to divide into similar target areas was constructed.
And (2): (S0) The image is imaged and input for each of the three colors R, G, B,
(S1) Display the RGB image on the color monitor,
(S2) The target areas 1, 2 and 3 are surrounded by a rectangular frame so as to be a composite of the monochromatic target areas 1, 2 and 3.
(S3) Surrounding the sampling pixel with the color of the area surrounded by the rectangular frame, setting the reference color R ₀ G ₀ B ₀ value of the target area,
(S4) The captured image is converted into a color distance image between the target area and the reference color R ₀ G ₀ B ₀ ,
(S5) Display the color distance of the region pixels surrounded by the rectangular frame on the histogram,
(S7) A color distance threshold suitable for target area division is set using the inflection point of the histogram as a guide,
(S8) The region surrounded by the rectangular frame is binarized by the color distance, displayed on the monitor,
(S9) Refer to the monitor to adjust the color distance threshold, repeat S2 to S9,
(S10) Calculate the area of the target area,
(S11) Calculate the gravity center position of the target area,
(S12) The image observation method is configured to register the figure number, the relative position in the group, the area, the area ratio, the reference color R ₀ G ₀ B ₀ , the color distance threshold value, and the target area frame as grouping figure properties.

Claim 3: (S0 ') The image is input for each of three colors R, G, and B,
(S1 ′) Display the RGB image on the color monitor,
(S2 ′) The captured and input image is converted into a color distance image between the reference color R ₀ G ₀ B ₀ of the target area,
(S4 ′) The region surrounded by the rectangular frame is binarized with the corrected color distance threshold,
(S5 ′) Calculate the target area,
(S6 ′) Calculate the target area centroid position,
(S7 ′) A target area layer is configured with the reference color R ₀ G ₀ B ₀ , area, and barycentric position as properties,
(S8 ") Search for the target area layer, superimpose the components of the registered figure group, create a figure pattern area layer,
(S9 ′) An image observation method for extracting a target figure pattern by multiplying a figure pattern or a target area layer and a color image input by imaging is configured.

Here, the present invention comprises a CCD color camera 100 that decomposes and outputs a captured RGB color image signal into an R image signal component, a G image signal component, and a B image signal component,
The set value storage units 511 to 51n respectively corresponding to the plurality of target regions 1 to n are provided, and the set value storage units 511 to 51n have R ₀ G ₀ B ₀ values and target threshold values for distinguishing the target regions. , The size of the target area is stored,
The R image signal component, the G image signal component, and the B image signal component are input, and the target regions 1 to n to which the R ₀ G ₀ B ₀ value is input from the corresponding target region setting value storage units 511 to 51n. The color distance calculation circuits 521 to 52n corresponding to the color distance calculation circuits 521 to 52n compare the RGB signal components of the pixels in the target regions 1 to n with the R ₀ G ₀ B ₀ reference value. Calculate the color distance using the color distance formula, convert the image of each target area to a color distance image,
The target area size set and stored in the set value storage units 511 to 51n and the color distance image output from the color distance calculation circuits 521 to 52n are input to generate a correction amount for the base threshold value and add it to the base threshold value. Histogram correction filters 531 to 53n for generating corrected color distance thresholds corresponding to the target areas 1 to n,
A binarization circuit 541 to 54n that inputs a color distance image and inputs a corrected color distance threshold value to generate a binary image of the target area is provided corresponding to the target areas 1 to n.
Layer output storage units 571 to 57n for storing the output binary image and the R ₀ G ₀ B ₀ value are provided corresponding to the target areas 1 to n,
A centroid position calculation unit 551 to 55n that inputs a binary image of the target area and calculates a centroid position corresponding to the target areas 1 to n,
Corresponding to the target areas 1 to n, area calculation units 561 to n that calculate the area by inputting a binary image of the target area,
The calculated barycentric position and area are output to the target region 1 to n in the layer configuration for each of the target regions 1 to n together with the directly output binary image and the R ₀ G ₀ B ₀ value, and stored in the target layer output storage units 571 to 57n. An image observation apparatus provided corresponding to the regions 1 to n was configured.

  In the conventional method in which the RGB signal is decomposed into the luminance signal and the chromaticity signal, and the luminance signal is image-processed to perform the target area division, three threshold values are required for the division. Then, when extracting the observation target area by dividing the surface image of the observation target into target areas having similar hues, the RGB signal of the surface image obtained by imaging with a CCD color camera is a numerical value called a color distance. And converting the surface image into a color distance image, binarizing the color distance image with the set color distance threshold value, and extracting the target area, thereby adopting a single color distance as the threshold value. A threshold is sufficient. The color distance threshold value can be set appropriately by creating a color distance histogram in each target area frame and using the inflection point of the histogram as a guideline and appropriately setting the color distance threshold value suitable for the target area extraction division process. .

The best mode for carrying out the invention will be described with reference to the embodiments shown in the drawings.
First, an outline of image observation according to the present invention will be described with reference to FIGS. 1 to 3. Reference numeral 1 denotes a reference numeral indicating the entire image observation apparatus. The image observation apparatus 1 includes an imaging device 10, a target area extraction unit 20, and an observation image monitor / target area setting unit 30. The imaging device 10 includes a CCD color camera 100 and images an observation target. The target region extraction unit 20 captures an 8-bit R (red) luminance signal, an 8-bit G (green) luminance signal, and an 8-bit B (blue) luminance signal output by imaging the observation target by the imaging device 10. This is a part where a target area extraction process is performed by inputting a signal called a RGB signal, which is a 24-bit signal of a three-dimensional vector obtained by combining three luminance signals. The observation image monitor / target area setting unit 30 is a part that performs target area setting processing. The target area setting process is performed prior to the target area extraction process. By configuring the image observation apparatus by combining the target area setting and the target area extraction, it is possible to classify the areas of the RGB color image without impairing the color difference component. Target area setting and target area extraction will be described in detail later.

Review the target area with reference to FIG. The observation target area 4 means a target to be observed and extracted from a surface image obtained by imaging a child as an observation target. The observation target area 4 includes a plurality of target areas 1, 2, and 3. Then, referring to FIG. 2, the target area is yellow-green as an example, the neighboring area that is the non-target area is gray as an example, and the color distance threshold is extracted from the area surrounded by the rectangular frame. It is defined as a threshold setting area or a target area frame.
Here, the color distance will be described. The observed image is observed with three vectors of R (red), G (green), and B (blue), and the distance amount with respect to the R ₀ G ₀ B ₀ value calculated for each pixel in the target region: √ [(R −R ₀ ) ² + (G−G ₀ ) ² + (B−B ₀ ) ² ] is called a color distance. The R ₀ G ₀ B ₀ value is a reference value of a sampling location that is a color reference in the target area. If the three vectors R, G, B of a certain pixel in the target area are R ₀ , G ₀ , B ₀ , respectively, the color distance of this pixel is 0, which is the same as the color at the sampling location Means color.

  Since the target area is originally a group of similar hues, it is displayed with a single RGB signal value, but the non-target area is not always displayed with a single RGB signal value. . In the target area and the non-target area, the color distance is calculated for each pixel, and the color image is converted into a color distance image based on the color distance. Then, both regions converted into the color distance image are classified by binarizing them using a set color distance threshold. That is, by binarizing the target area and the non-target area with the color distance threshold, a portion exceeding the color distance threshold is converted to black and displayed, while a portion not exceeding the color distance threshold is converted to white and displayed. The

The details of the division and division of the target area and the non-target area in the image observation will be specifically described with reference to FIGS. 4, 5, and 6. FIG. 4 is a diagram for explaining target area extraction of a target graphic pattern, and FIG. 5 is a diagram for explaining an embodiment of the image observation apparatus. FIG. 6 is a diagram showing an operation sequence in which target area setting and target area extraction are integrated. The target area setting is performed by sequentially following the steps in the left chain line frame, and the target area extraction is performed in the right chain line frame. Follow the steps in sequence.
Here, the definition of “region” in region extraction of the target graphic pattern is summarized in Table 1.

以下、目標領域の抽出を図４を参照して説明する。
目標領域１、２、３の見当をつけるには、目標図形パターンが含まれる図示される通りの画像をモニタ画面上に表示する。
観測目標領域４を構成する要素の候補として、図４において黄色の四角の枠で囲んだ目標領域１〜目標領域３を設定する。この設定は、具体的には、モニタ画面上においてマウスを操作することにより黄色の四角の枠で包囲設定する。

Hereinafter, the extraction of the target area will be described with reference to FIG.
In order to register the target areas 1, 2, and 3, an image as shown in the figure including the target graphic pattern is displayed on the monitor screen.
As candidate elements constituting the observation target area 4, target areas 1 to 3 surrounded by a yellow square frame in FIG. 4 are set. Specifically, this setting is performed by setting a siege with a yellow square frame by operating the mouse on the monitor screen.

The meaning of the target area 1 to the target area 3 is a bicycle tire and a red outerwear, and the definitions are a yellow area and a red area.
For these target areas 1 to 3, sampling locations that are supposed to serve as color references in the target areas are set by operating the mouse, and each reference R ₀ G is set for each of the target areas 1 to 3 from these sampling locations. ₀ B Read ₀ value. Here, a place indicated by a small square indicates a sampling place. A norm operation is performed for each pixel based on the RGB value of the image input and the reference R ₀ G ₀ B ₀ value to convert the target area into a color distance image for each target area, and an image is created for each layer. Here, the meaning of the layer (or hierarchy) will be described as follows.

  Referring to FIG. 9, an area having a certain set color distance, that is, an image for each of red, orange, yellow, and yellowish green is created, and this image is binarized. Are referred to as layers (or hierarchies) 1, 2, 3, and 4. Layer 1 is a color distance binarized image for the red region, and an image is obtained corresponding to only the red region existing in the center. The layer 2 is a color distance binarized image for the orange region, and an image is obtained corresponding to only the orange region surrounding the red region. Layer 3 is a color distance binarized image for the yellow region, and an image is obtained corresponding to only the yellow region surrounding the orange region. The layer 4 is a color distance binarized image with respect to the yellow-green region, and an image is obtained corresponding to only the yellow-green region surrounding the outside of the yellow region.

Referring to FIG. 10, layer 1 is a color distance binarized image with respect to a yellow region, and an image is obtained corresponding to only a triangular region. Layer 2 is a color distance binarized image for the red region, and an image is obtained corresponding to only the circular region. Layer 3 is a color distance binarized image for a blue region, and an image is obtained corresponding to only a rectangular region. Layer 4 is a color distance binarized image for a yellow-green region, and an image is obtained corresponding to only the yellow-green region surrounding the outside of the yellow, red, and blue regions.
In FIG. 10A, the yellow, red, blue, and yellow-green regions having different color distances are binarized into color distances for each layer and layer as the extracted region images. The number, position, area relationship, and conditions of the extraction regions are recognized with reference to FIG.

A color distance threshold is set, and a target area is extracted by binarization and normalization of the color distance image based on the color distance threshold for each layer. That is, by binarizing and normalizing the target area and the non-target area by the color distance threshold, a portion exceeding the color distance threshold is displayed in black, while a portion not exceeding the color distance threshold is displayed in white. Extract the target area.
The shape of the target area is compared using three parameters: the barycentric position, the area, and the reference R ₀ G ₀ B ₀ value.
The target graphic pattern is extracted by comparing the relative position between the target areas, the area of each target area, and the reference R ₀ G ₀ B ₀ value of each target area with the target graphic pattern across the layers. An example of the relative position is shown in FIG. FIG. 7 is a diagram showing an example of a condition for extracting a target graphic pattern by a group of target areas. The matching pattern shown in Table 2 can be defined by combining conditions for extracting a target graphic pattern by a group of target areas. (S7) will be described more specifically later.

図４において、目標領域１、２の意味は自転車のタイヤであり、目標領域３の意味は赤い上着であり、その定義は黄色の領域と赤色の領域である。これら黄色の四角の枠で囲んだ目標領域１〜目標領域３を青枠で囲んで一つのグループの目標図形観測パターンとして抽出する。「横に並んだ目標領域１、２、更にその中間上部に配置する目標領域３」という定義の下で青枠で囲んで一つのグループの観測目標領域４としてグループ化することで、この定義を満足するか否かを判断して、目標領域１〜目標領域３を一括して抽出することができる。

In FIG. 4, the meanings of the target areas 1 and 2 are bicycle tires, the meaning of the target area 3 is a red jacket, and the definitions are a yellow area and a red area. The target area 1 to the target area 3 surrounded by the yellow square frame are extracted as a target figure observation pattern of a group surrounded by a blue frame. By grouping it as an observation target area 4 of one group surrounded by a blue frame under the definition of "target areas 1 and 2 arranged side by side, and target area 3 arranged in the middle upper part", this definition It is possible to extract the target area 1 to the target area 3 collectively by determining whether or not the user is satisfied.

An embodiment of the image observation apparatus will be described with reference to FIG. This is to extract a target area with a single reference value, and input the R ₀ G ₀ B ₀ value of each target area, the base threshold value for distinguishing each target area from the non-target area, and the size of the target area The R ₀ G ₀ B ₀ value, the binary image of the target area, the barycentric position, and the area of the target area extracted as are output in a layer configuration for each target area. This will be specifically described below.
100 is a CCD color camera, 511 to 51n are set value storage units, 521 to 52n are color distance calculation circuits, 531 to 53n are histogram extraction filters, 541 to 54n are binarization circuits, 551 to 55n are gravity center position calculation circuits, Reference numerals 561 to 56n denote area calculation circuits, 571 to 57n denote layer output storage units, and 58 denotes an extracted image output unit.

An RGB color image signal captured by the CCD color camera 100 is decomposed into an R image signal component, a G image signal component, and a B image signal component and output. Corresponding set value storage units 511 to 51n are provided in the target areas 1 to n. The set value storage units 511 to 51n store the R ₀ G ₀ B ₀ value of each target area, the base threshold value for distinguishing the target area, and the size of the target area.
The R image signal component, the G image signal component, and the B image signal component are input to the color distance calculation circuits 521 to 52n corresponding to the target regions 1 to n, and the corresponding target region setting value storage units 511 to 511 are used. The R ₀ G ₀ B ₀ value is input from 51n. In these color distance calculation circuits 521 to 52n, the RGB signal components of each pixel in each target region 1 to n are compared with the R ₀ G ₀ B ₀ reference value, and the color distance is calculated by the previous color distance calculation formula, The image of each target area is converted into a color distance image.

Reference numerals 531 to 53n denote histogram extraction filters respectively corresponding to the target areas 1 to n. Reference numerals 541 to 54n denote binarization circuits.
By the way, the base threshold supplied from the set value storage units 511 to 51n is subjected to the following processing prior to being input to the binarization circuits 541 to 54n. When setting the target area, a base value of the color distance threshold is set. In order to set the base value of the threshold, first, the target area occupies the main ratio, and a frame of the target area inscribed in the target area is set. That is, the size of the frame of the target area used for calculating the difference for raising the base threshold is set. Then, the inside of the target area frame is converted into a color distance image, the histogram of the color distance distribution shown in FIG. 8 is taken, and the minimum value or inflection point of the histogram is set as the base threshold value for the entire target area frame. Each of the histogram extraction filters 531 to 53n receives the target area size set and stored in the set value storage units 511 to 51n and the color distance image output from the color distance calculation circuits 521 to 52n, and corrects the color distance threshold. Generate quantity. A color distance corrected by generating a difference value of a target area between the target area of the RGB color image and other areas as a correction amount and adding this to the base threshold supplied from the set value storage units 511 to 51n A threshold value is generated, and the target area is binarized and input to the binarization circuit 541, whereby the target area and other areas are divided and distinguished. This is because the color distance threshold value for binarizing the target area and the other non-target areas is set as a small value close to the target area as a base value, and the target area and other areas This is because when the non-target area has a large distance in the color distance, the binarization can be stabilized by setting the distance higher corresponding to the distance.

The binary image of the target area generated by the binarization circuits 541 to 54n is output to the corresponding layer output storage units 571 to 57n and stored therein. Then, the binary image of the target area is supplied to the corresponding centroid position calculation units 551 to 55n, the centroid position is calculated, output to the corresponding layer output storage units 571 to 57n, and stored therein. Further, the binary image of the target area is supplied to the corresponding area calculation units 561 to n, the area is calculated, output to the corresponding layer output storage units 571 to 57n, and stored therein. The layer output storage units 571 to 57n are further supplied with R ₀ G ₀ B ₀ values from the corresponding set value storage units 511 to 51n. As described above, for the binary image of the binarized target area, the area and the centroid position are calculated for each target area, and the calculated centroid position and area are directly output as a binary image and R ₀ G ₀ B _0. Along with the value, it is output and stored in a layer configuration for each of the target areas 1 to n.

  Here, referring to FIG. 8, this is a color distance histogram between the target area and the non-target area. The “area ratio” on the vertical axis is a value obtained by dividing the area of the target region by the area of the rectangular frame. It is understood that, under a condition having a clear outline and a small color distance, the peak is clearly separated as shown by the blue curve, and the threshold setting is easy. On the other hand, it is understood that it is difficult to set a clear threshold when the boundary is blurred. Under this condition, the size of the binarized image changes depending on the threshold value. As described above, when the color distance around the target area is not constant, the binarization of the target area can be stabilized by raising the difference of the target area of the color distance image to the base threshold value.

FIG. 6 is a diagram showing an operation sequence in which target area setting and target area extraction are integrated. The target area setting is performed by sequentially following the steps in the left chain line frame, and the target area extraction is performed in the right chain line frame. Follow the steps in sequence. Hereinafter, an operation sequence for setting a target area of an observation area will be described with reference to FIGS. 4 and 6.
(S0) A CCD color camera picks up and inputs an original image composed of three images of an R image, a G image, and a B image to be observed.
(S1) The color original image including the target area is displayed on the monitor screen.

(S2) The target areas 1, 2, and 3 are surrounded by a rectangular frame on the monitor screen, and the graphic pattern of the observation area is divided into a complex of single-color target areas 1, 2, and 3. This is set as the target frame.
(S3) On the monitor screen, the sampling pixel that is the reference for the color in each target area is enclosed, and the R ₀ G ₀ B ₀ value of the reference pixel is set.
(S4) Each pixel in each target area frame is converted into a color distance between the R ₀ G ₀ B ₀ values for each pixel, and the inside of each target area frame is converted into a color distance image.
(S5) A color distance histogram in each target area frame is created.

(S7) A color distance threshold suitable for the target area extraction division process is set using the inflection point of the histogram as a guide.
(S8) Each target area frame is binarized by comparing with the color distance threshold value, and the binarized image is displayed on the monitor.
(S9) Referring to the monitor screen, the color distance threshold is adjusted to check whether the threshold is appropriate. If the extracted target area does not accurately extract the target area, the process returns to (S2), the setting of the target area frame and the setting of the color distance threshold are reset, and the binarized image is optimized. Repeat the above steps until

(S10) If the target area is accurately extracted, the area of the target area is calculated.
(S11) The gravity center position of the target area is calculated.
(S12) The target area is registered. The R ₀ G ₀ B ₀ value, the color distance threshold value, the target area frame, and the target area of the target area are registered. The figure registration property includes the figure number to be grouped, the relative position in the group, the area, the area ratio, the absolute position on the set drawing, and the R ₀ G ₀ B ₀ value of the target area.
Here, the target area extraction process will be described by sequentially following the steps in the chain line frame on the right side of FIG.

(S0 ′) The CCD color camera 100 captures and inputs an original image composed of three images, an R image, a G image, and a B image to be observed.
(S1 ′) Color distance conversion calculation between R ₀ G ₀ B ₀ values for each pixel in each target area frame √ [(R−R ₀ ) ² + (G−G ₀ ) ² + (B−B ₀ ) _Perform ² ].
(S2 ′) A color distance image of each target area is created.
(S3 ′) A difference image is calculated with the size of the target area to create a corrected image of the color distance threshold.
(S4 ′) A binarized image is created by binarizing the target area by comparing with the color distance threshold corrected for each pixel of the target area.

(S5 ′) The area of the binarized image of the target area is calculated.
(S6 ′) The gravity center position of the target area is calculated.
(S7 ′) The target region layer is configured using the R ₀ G ₀ B ₀ value, area, and barycentric position as properties.
(S8 ′) The target area layer is searched and the constituent elements of the registered graphic group are overlapped to create a graphic pattern area image layer.
(S9 ') The figure pattern or the target area layer and the color image are multiplied to extract the target figure pattern.

  FIG. 7 is a diagram showing an example of conditions for extracting the observation target target area 4 in groups of the target areas 1, 2, and 3. The matching pattern shown in Table 2 is defined by combining the conditions for extracting the observation target region 4 by the group of the target regions 1, 2, and 3. In FIG. 7, Condition 1 is that there are three target areas and three extraction areas. If the number of regions is the same, it can be determined that matching is performed according to matching pattern 1 in Table 2. The number of target areas and extraction areas that satisfy the conditions 1, 2, and 3 are the same, the areas of the target area and the extraction areas are equal, and the relative positional relationship between the target areas and the relative position of the extraction areas When the relationship is equal, it can be determined that matching is performed according to the matching pattern 4 in Table 2.

The present invention is preferably applied in the field of measuring defects of semiconductor materials and the like and the field of measuring microorganisms under a microscope. In addition, it also demonstrates highly efficient target area extraction capabilities in environmental monitoring and production line monitoring. In addition, smoke, greening, forest CO ₂ monitoring, acid rain monitoring, other environmental monitoring, food foreign matter monitoring, food measurement monitoring by shape and color, and continuous monitoring monitoring can be made intelligent and applied to security monitoring. .

The figure explaining the outline | summary of an image observation apparatus. The figure explaining a target area | region. The figure explaining how to extract the RGB signals as a single observation area by combining a plurality of target areas. The figure which shows the target area frame containing a target area. The figure explaining the Example of an image observation apparatus. The figure which shows the operation | movement sequence which integrated target area | region setting and target area | region extraction. The figure which shows the example of the conditions which extract a target figure pattern in the group of a target area | region. Color distance histogram between target area and non-target area. The figure explaining a layer. The figure explaining another layer.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 CCD color camera 511-51n Setting value memory | storage part 521-52n Color distance arithmetic circuit 531-53n Histogram extraction filter 541-54n Binary circuit 551-55n Center of gravity position arithmetic circuit 561-56n Area arithmetic circuit 571-57n Layer output storage Part
58 Extracted image output unit

Claims (4)

  The RGB signal of the surface image obtained by imaging with a CCD color camera is converted into a color distance, the target area is binarized with a single color distance threshold, and the target area having a similar hue on the surface image of the observation target An image observation method characterized by dividing the image into two. (S0) An image is input for each of the three colors R, G, and B, and
(S1) Display the RGB image on the color monitor,
(S2) The target areas 1, 2 and 3 are surrounded by a rectangular frame so as to be a composite of the monochromatic target areas 1, 2 and 3.
(S3) Surrounding the sampling pixel with the color of the area surrounded by the rectangular frame, setting the reference color R ₀ G ₀ B ₀ value of the target area,
(S4) The captured image is converted into a color distance image between the target area and the reference color R ₀ G ₀ B ₀ ,
(S5) Display the color distance of the region pixels surrounded by the rectangular frame on the histogram,
(S7) A color distance threshold suitable for target area division is set using the inflection point of the histogram as a guide,
(S8) The region surrounded by the rectangular frame is binarized by the color distance, displayed on the monitor,
(S9) Refer to the monitor to adjust the color distance threshold, repeat S2 to S9,
(S10) Calculate the area of the target area,
(S11) Calculate the gravity center position of the target area,
(S12) Register the figure number, the relative position in the group, the area, the area ratio, the reference color R ₀ G ₀ B ₀ , the color distance threshold value, and the target area frame that group the figure properties.
An image observation method characterized by that. (S0 ′) The image is input for each of three colors R, G, and B, and
(S1 ′) Display the RGB image on the color monitor,
(S2 ′) The captured and input image is converted into a color distance image between the reference color R ₀ G ₀ B ₀ of the target area,
(S4 ′) The region surrounded by the rectangular frame is binarized with the corrected color distance threshold,
(S5 ′) Calculate the target area,
(S6 ′) Calculate the target area centroid position,
(S7 ′) A target area layer is configured with the reference color R ₀ G ₀ B ₀ , area, and barycentric position as properties,
(S8 ") Search for the target area layer, superimpose the components of the registered figure group, create a figure pattern area layer,
(S9 ′) The target graphic pattern is extracted by multiplying the graphic pattern or the target area layer and the color image captured and input,
An image observation method characterized by that. A CCD color camera that decomposes and outputs the captured RGB color image signal into an R image signal component, a G image signal component, and a B image signal component;
A set value storage unit corresponding to each of the target areas 1 to n is provided, and the R ₀ G ₀ B ₀ value of each target area, a base threshold value for distinguishing the target area, and the size of the target area are stored in the set value storage section. Stored,
The R image signal component, the G image signal component, and the B image signal component are input, and the R ₀ G ₀ B ₀ value is input from the set value storage unit of the corresponding target region. A color distance calculation circuit is provided. In these color distance calculation circuits, the RGB signal component of each pixel in each target area 1 to n is compared with the R ₀ G ₀ B ₀ reference value, and the color distance is calculated by the color distance calculation formula. Calculate, convert the image of each target area into a color distance image,
The target area size set and stored in the set value storage unit and the color distance image output from the color distance calculation circuit are input to generate a base threshold correction amount, which is added to the base threshold and corrected color distance A histogram extraction filter for generating a threshold is provided corresponding to the target areas 1 to n,
A binarization circuit that inputs a color distance image and inputs a corrected color distance threshold to generate a binary image of the target area is provided corresponding to the target areas 1 to n,
A layer output storage unit for storing the output binary image and the R ₀ G ₀ B ₀ value is provided corresponding to the target areas 1 to n,
A centroid position calculation unit that inputs a binary image of the target area and calculates a centroid position corresponding to the target areas 1 to n,
Corresponding to the target areas 1 to n, an area calculation unit that calculates the area by inputting a binary image of the target area is provided.
The calculated barycentric position and area are output in a layer configuration for each of the target regions 1 to n together with the binary image and R ₀ G ₀ B ₀ value that are directly output, and a layer output storage unit that stores this is output to the target regions 1 to 1 corresponding to n,
An image observation apparatus characterized by that.

Images