KR20200017583A - Apparatus and method for determining the state of area including water in an image using color labeling - Google Patents

Abstract

The present invention relates to a method for determining the state of a water-containing region of a video image using color labeling. More specifically, according to the present invention, the method for determining the state of a water-containing region of a video image using color labeling comprises the steps of: dividing regions according to a plurality of preset divisions based on a water-containing state in each original image, and labeling the same with a preset color corresponding to the divided regions to generate a plurality of labeling images labeled with different colors for each area, in the plurality of original images including, as a part constituting the image, a region visually divided and recognizable according to the water-containing state such as a region filled with water, a region wet with water, a dry region, and an area of no interest; analyzing an image feature by using a plurality of patterned masks on the labeling image labeled with at least one color, obtaining feature pattern components according to a preset feature item corresponding to each of the patterned masks, and generating a feature map based on the feature item and the feature pattern components; analyzing the image features by using the patterned mask for an identification target image requiring the identification of the water-containing state, extracting the feature pattern components, and calculating a comparison value according to comparison with the feature map; and determining the water-containing state of the identification target image based on the preset division and the comparison value. Therefore, accuracy can be increased when determining the water-containing state.

Description

Apparatus and method for judging the water-containing area of an image image using color labeling {APPARATUS AND METHOD FOR DETERMINING THE STATE OF AREA INCLUDING WATER IN AN IMAGE USING COLOR LABELING}

The present invention relates to an apparatus and method for determining a water-containing region state of an image image using color labeling. More particularly, the present invention relates to a color for easily determining a water-containing state in an image image using color labeling. An apparatus and method for judging a water-containing region state of an image image using labeling are provided.

Photographs or video images can be stored as binary coded digital values, and attempts have been made to determine the type or state of objects in photographic or video images through image processing.

In a photographic or video image, when there is an area filled with water, a wet area, a dry area, an uninterested area, etc., identifiable by human vision, the person visually includes water contained in the picture or video image. It is easy to identify the state of the water and the area containing water, but the boundary between the area containing water is often unclear, and the characteristics of the area containing water are not regular, so that the water is removed by the image processing technique. There is a limit that it is difficult to easily grasp the area to be included.

In addition, water-filled areas, such as rivers, valleys, and rivers, change in water level over time, and because water flows from upstream to downstream, the characteristics of the water-filled area change over time. There is a limitation that it is not easy to determine the area filled with water according to the treatment technique. In addition, the part that forms the boundary between the cold and dry areas is mostly wetted with water unless it is an artificial structure, and the boundary varies depending on the level of the cold area and the flow of water. Due to the difficulty in extracting the boundary value, there is a limitation that it is not easy to discriminate the area according to the existing image processing technique.

KR 10-0731692 B1 KR 10-1768521 B1 JP 2015-032308 A

An object of the present invention is to provide a color to images that include visually distinguishable and recognizable areas as part of an image, including water filled areas, wet areas, dry areas, uninterested areas, and the like. Apparatus for judging a water-containing area of an image image using color labeling to determine a state of containing water in a video image by using feature pattern components obtained from the images to which color labeling is applied and color classification; To provide a way.

The object of the present invention is not limited to the above-mentioned object, and other objects which are not mentioned will be clearly understood from the following description.

According to an aspect of the present invention, a method for determining a water-containing region state of an image image using color labeling includes water, such as a region filled with water, a region wet with water, a dry region, and an uninterested region. In a plurality of original images that include visually distinguishable and recognizable areas according to states as part of constituting an image, the areas are divided according to a plurality of predetermined divisions based on a state of containing water in each original image. And generating a plurality of labeling images labeled with different colors for each area by labeling the predetermined color corresponding to each divided area, and using a plurality of patterned masks for the labeling image labeled with at least one color. Image features, and preset features corresponding to each of the patterned masks. Acquiring the feature pattern components according to the item, creating a feature map based on the feature item and the feature pattern components, and using a patterned mask for the identification image for which identification of a state containing water is required Analyzing image features, extracting feature pattern components, calculating a comparison value according to comparison with a feature map, and determining a state including water of the image to be identified based on a predetermined division and a comparison value; do.

According to another aspect of the present invention, an apparatus for determining a water-containing region state of an image image using color labeling includes: in the original images including a visually distinguishable and recognizable region as part of an image according to a state including water. A labeling image providing unit which provides a plurality of labeling images by dividing a plurality of original images according to a predetermined division and labeling the plurality of original images with a predetermined color corresponding to the divided region, and identification of a state including water Analyze the image feature using an identification image providing unit for receiving an input of the identification image to be received from the user or the outside, and a plurality of patterned masks for the labeling image or the identification image labeled with at least one color, and the pattern According to a feature item corresponding to each of the masked masks A feature map that generates a feature map based on a feature pattern component acquirer for obtaining pattern components, patterned masks used in the feature pattern component acquirer, and feature pattern components obtained using the patterned masks. And a comparison value calculator for comparing the feature pattern components of the identification target image with the feature map created by the feature map generator to calculate a comparison value.

According to the present invention, a state of containing water, such as a region filled with water, a region wet with water, a dry region, an uninterested region, and the like in a video image based on feature pattern components and feature maps extracted from labeled images. There is an advantage that it can be easily determined according to the predetermined division.

In addition, according to the present invention, there is an advantage that it is possible to increase the accuracy in learning the labeled images to determine the state containing water in the image more.

Further, according to the present invention, since a convolution operation is performed on a pixel basis of an image, a probabilistic output value can be calculated in distinguishing a state including water in an image image.

1 is a block diagram illustrating an apparatus for determining a water-containing region state of an image image using color labeling according to an exemplary embodiment of the present invention.
2 is a flowchart illustrating a method of determining a water-containing region state of an image image using color labeling according to another exemplary embodiment of the present invention.
3 is an exemplary diagram for explaining an original image of a river.
4 is an exemplary view for explaining a labeling image according to an embodiment of the present invention.
5 is an exemplary diagram for describing a patterned mask according to an embodiment of the present invention.
6 is an exemplary view for explaining an image to be identified according to an embodiment of the present invention.
7 is an exemplary view illustrating a region determined as an area including water in an image to be identified according to an embodiment of the present invention.
8 is a diagram illustrating overlapping and displaying a region including water in an image to be identified according to an embodiment of the present invention.

Advantages and features of the present invention, and methods for achieving them will be apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various forms, only the embodiments are to make the disclosure of the present invention complete, and those skilled in the art to which the present invention pertains It is provided to fully inform the person of the scope of the invention. Meanwhile, the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Hereinafter, an apparatus for determining a water containing area state of an image image using color labeling and a method for determining a water containing area state of an image image using color labeling will be described with reference to FIGS. 1 and 2.

1 is a view showing an apparatus for determining a water containing area of an image image using color labeling according to an embodiment of the present invention, and FIG. 2 is a water containing area of an image image using color labeling according to another embodiment of the present invention. It is a flowchart showing a state determination method.

The apparatus 10 for determining a water-containing region state of an image image using color labeling according to an embodiment of the present invention may include a labeling image providing unit 100, an identification target image providing unit 200, and a pattern mask generating unit ( 300, the feature item setting unit 400, the feature pattern component obtaining unit 500, the feature map preparation unit 600, the learning rate setting unit 700, the comparison value calculating unit 800, and the discrimination The unit 900 is included.

In the apparatus and method for judging the water-containing area of an image image using color labeling according to an embodiment of the present invention, first, according to a state including water, such as a region filled with water, a region wet with water, a dry region, an uninterested region, etc. In a plurality of original images that include visually distinguishable and recognizable areas as part of an image, the regions are divided and divided according to a plurality of predetermined divisions based on a state of containing water in each original image. In operation S100, a plurality of labeling images labeled with different colors are generated by labeling a predetermined color corresponding to an area.

Hereinafter, a labeling image according to an exemplary embodiment of the present invention will be described with reference to FIGS. 3 and 4, which are examples of a labeling image corresponding to the original image of FIG. 3.

 In FIG. 3, since the original image is an image of a river, the image is divided into visually distinguishable and recognizable areas according to a state including water, such as a region filled with water, a region wet with water, a dry region, and an uninterested region. It is included as a part, and includes a region filled with water corresponding to a stream in which the water flows, 901, and a region wetted by a water level fluctuation adjacent to the boundary between the stream and the ground to 903, 901, or 903. Dry areas that are not separated can be divided to 905.

In this case, as shown in FIG. 4, each divided area may be labeled with a predetermined color corresponding to a region filled with water, a region wet with water, a dry region, and an uninterested region. For example, area 901, which is full of water, is overlaid and labeled to have a predetermined transparency in blue color, and area 903, which is wet, is overlaid and labeled to have a preset transparency in green color, and the remaining area is The region 905 may be labeled by overlaying the red color to have a preset transparency and labeling or not labeling the original image.

Here, the predetermined color may be a color selected from among colors according to the RGB color model. In FIG. 4, for convenience of description, different identification marks (for example, x, +,-) for each color are used instead of the color to distinguish the area filled with water, the area wetted with water, the dry area, and the uninterested area. It was shown.

If the original image is labeled using color by dividing the area according to the state of containing water, there is an advantage that the division of the area according to the state of containing water and the boundary between the areas become clear. In addition, when labeling using color, in addition to the color information in the pixel information constituting the original image, information on other image characteristic components (for example, a line component, a pattern component, etc.) does not change, and thus even after the labeling process. The advantage is that the effect is similar to that of analyzing the image characteristic of the original image.

The plurality of labeling images may be generated from one original image or a plurality of original images, and may be to generate a plurality of labeling images classified by area or color according to a state including water.

For example, in a plurality of different original images photographing a stream, a plurality of labeling images may be generated in which only a wet area is labeled in green, and a water-filled region may be generated from a single original image of a stream. A plurality of labeling images may be generated by labeling a blue, water-wet area in green, and a dry area in red.

On the other hand, the plurality of labeling images may be generated in the water-containing region state determination device 10 of the image image using the color labeling or may be provided from the outside.

The labeling image providing unit 100 receives a plurality of original images, divides the regions according to a predetermined division, and labels them in a predetermined color corresponding to the divided regions to generate a plurality of labeling images labeled in different colors for each region. It may be.

In addition, the labeling image providing unit 100 may receive a plurality of labeled image labels from a user or may be provided from the outside and stored in a database form.

The identification object image providing unit 200 receives an input from the user or the outside from the user or the outside to provide the characteristic pattern component obtaining unit 500 to identify the image including the water.

Here, the identification target image may be an image as any one of a plurality of images constituting an image frame.

The identification target image providing unit 200 stores and manages a plurality of identification target images to identify the feature pattern component acquirer 500 or the feature map generator 600 sequentially or sequentially. It may be to provide target images.

The pattern mask generator 300 generates a mask having a predetermined size and generates a patterned mask by applying a binary value to each pixel constituting the mask to form a pattern.

5 illustrates an example of a patterned mask in which a “Z” pattern having a size of 5 × 5 (Fixel X Fixel) is formed, and each pixel may be given a value of '1' or '0' as a binary value. .

The pattern mask generator 300 may generate a plurality of pattern masks having different patterns.

The feature item setting unit 400 sets a feature item according to a shape of a pattern corresponding to each of the masks patterned according to a user's input.

Here, the feature item may be an item having a pattern, a letter, a number, etc., which symbolizes the shape of a pattern, and may be an item having a geometric element such as a line, a figure, a curve, or the like. In addition, the feature item may be an item having a feature according to image analysis.

The feature pattern component acquirer 500 analyzes an image feature using a plurality of patterned masks for a labeling image or an image to be labeled with at least one color, and presets corresponding to each of the patterned masks. The feature pattern components are acquired according to the feature item according to the feature item or the feature item setting unit 400 (S200 and S400).

Meanwhile, the feature pattern component acquirer 500 features the pixel values constituting the labeling image and pixel values constituting at least one patterned mask among the patterned masks according to a result of the convolution operation. It may be to obtain the pattern components.

The feature map generator 600 may include the feature items of the feature item setting unit 400, the feature pattern components obtained by using the patterned masks and the patterned masks used in the feature pattern component acquirer 500. A feature map is created based on S300.

The feature map generator 600 may obtain feature pattern components using a plurality of patterned masks for the plurality of labeling images, and organize the feature maps by feature items to create a feature map.

On the other hand, since the plurality of labeling images used when the feature map is created are labeled by dividing the area according to the state of containing water, the feature map generator 600 is obtained from the labeling images labeled with the same color. The feature map may be generated for each color or for each region by using the feature pattern components.

In this case, the feature pattern component acquirer 500 may include a first labeling image labeled with at least one color and a second labeling image different from the first labeling image labeled with the same color as the color used for labeling the first labeling image. The image feature may be analyzed using patterned masks for, and the feature pattern components may be obtained according to a predetermined feature item corresponding to each of the patterned masks.

The feature map generator 600 may repeatedly perform the process of acquiring the feature pattern components according to the feature item using a plurality of labeling images, and may modify and change the previously created feature map by using a preset learning model.

The feature map generator 600 is a first value that is a result of convolution of pixel values constituting at least one patterned mask among respective pixel values constituting the first labeling image and the patterned masks. And generating a feature map for each labeling color based on a second value obtained by multiplying a predetermined learning rate by a convolution operation of pixel values constituting the turned mask with the feature pattern components obtained from the second labeling image. Can be.

The learning rate setting unit 700 sets a learning rate according to a user input, and the feature map preparing unit 600 may create a feature map by reflecting the learning rate set by the learning rate setting unit 700.

Reflecting the learning rate when creating the feature map, the feature map components can be efficiently created using the feature pattern components obtained initially, while the feature map components can be updated to reflect the weights of the feature pattern components obtained later. As feature items and feature pattern components accumulate, it is possible to repeat the learning and create a feature map according to the repeated learning, thereby increasing the accuracy.

The comparison value calculator 800 calculates a comparison value by comparing the feature pattern components of the identification target image requiring identification of a state including water with the feature map created by the feature map creator 600 (S500). .

The comparison value calculator 800 may calculate a ratio value of pixels that match within a predetermined range between the feature pattern components and the feature map of the identification image.

Meanwhile, the comparison value calculating unit 800 may calculate a comparison value according to the degree of matching of the feature pattern components and the number of feature items and feature pattern components included in the feature map with respect to the image to be identified. have.

The determination unit 900 determines a state including water of the identification target image based on a predetermined division and a comparison value (S600). For example, the comparison value calculator 800 may determine a state including water of the image to be identified as a 30% area filled with water, 45% area wetted by water, 25% area dry, and the like. On the other hand, if there are overlapping areas in each of the divided areas, the sum of the comparison values for the divided areas may be 100 or more than 100%.

According to an embodiment of the present invention, the region including water in the identification target image illustrated in FIG. 6 according to the result determined by the determination unit 900 may be displayed as shown in FIG. As described above, an area including water may be overlapped and displayed in the image to be identified.

The apparatus 10 for determining a water containing area of an image image using color labeling according to an embodiment of the present invention may include an image to be identified, a region filled with water, a region wet with water, a dry region, an uninterested region, and a determination unit 900. Display means (not shown) for displaying a comparison value, etc .;

According to an apparatus and method for determining a water containing area state of an image image using color labeling, a plurality of labeling images according to an exemplary embodiment of the present invention use a plurality of labeling images that are labeled in different colors for each area according to a water containing state. The image characteristic components according to the water-containing state are extracted using a patterned mask, and a feature map is created based on the image characteristic components extracted using different patterned masks. In addition, as the number of patterned masks for extracting the image characteristic components increases, and the attempt to extract the image characteristic components is repeated, the accuracy of the water inclusion region state determination using the feature map may increase.

According to an apparatus and method for determining a water-containing area state of an image image using color labeling, a plurality of labeling images according to an exemplary embodiment of the present invention are extracted from a feature map and a target image generated using the plurality of labeling images. Since the state of containing water of the image to be identified is determined by comparing the pattern components, the state of containing water in the image to be identified may be easily quantitatively and qualitatively determined according to a predetermined classification.

Those skilled in the art will appreciate that the present invention can be embodied in other specific forms without changing the technical spirit or essential features of the present invention. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive. The scope of the invention is indicated by the following claims rather than the above description, and it should be construed that all changes or modifications derived from the claims and their equivalents are included in the scope of the invention.

10: apparatus for judging the water containing area of an image image using color labeling,
100: labeling image providing unit,
200: identification target image providing unit,
300: pattern mask generation unit,
400: feature item setting unit,
500: feature pattern component acquisition unit,
600: feature map creation unit,
700: learning rate setting unit,
800: comparison value calculation unit,
900: Determination unit.

Claims (12)

In a plurality of original images that include visually distinguishable and recognizable areas according to a state of containing water as part of constituting an image, a plurality of preset divisions based on a state of containing water in each original image are included. Generating a plurality of labeling images labeled with different colors for each region by dividing the regions and labeling the regions with a predetermined color corresponding to the separated regions;
Analyze image features using a plurality of patterned masks on at least one color-labeled labeling image, obtain feature pattern components according to predetermined feature items corresponding to each of the patterned masks, and Creating a feature map based on a feature item and the feature pattern components;
Analyzing an image feature using a patterned mask on an image to be identified that requires identification of a state including water, extracting feature pattern components, and calculating a comparison value according to comparison with the feature map; And
Determining a state including water of the identification target image based on the predetermined division and the comparison value;
Method for determining the water containing area of the image image using color labeling comprising a. The method of claim 1,
Creating the feature map,
Using the patterned masks for a first labeling image labeled with at least one color and a second labeling image that is different from the first labeling image labeled with the same color used to label the first labeling image. Analyzing image features and acquiring feature pattern components according to predetermined feature items corresponding to each of the patterned masks
A method for judging the water containing area of an image image using phosphor color labeling. The method of claim 2,
Creating the feature map,
Acquiring the feature pattern components according to a convolution operation of respective pixel values constituting the first labeling image and pixel values constituting at least one patterned mask among the patterned masks;
A method for judging the water containing area of an image image using phosphor color labeling. The method of claim 2,
Creating the feature map,
Creating a feature map for each labeling color by reflecting a predetermined learning rate in the feature pattern components obtained from the first labeling image and the second labeling image;
A method for judging the water containing area of an image image using phosphor color labeling. The method of claim 2,
The step of creating the feature map
A first value that is a result of a convolution operation of respective pixel values constituting the first labeling image and pixel values constituting at least one patterned mask among the patterned masks, and a second labeling image Creating a feature map for each labeling color based on a second value obtained by multiplying the predetermined learning rate by a convolution operation of the feature pattern components obtained from the pixel values constituting the patterned mask.
A method for judging the water containing area of an image image using phosphor color labeling. The method of claim 2,
The calculating step,
Calculating a ratio value of pixels matched within a predetermined range between the feature pattern components and the feature map for the identification image;
A method for judging the water containing area of an image image using phosphor color labeling. In original images including a visually divided and recognizable area according to a state of containing water as part of an image, a plurality of original images are divided into areas according to a predetermined division and corresponding to the divided areas. A labeling image providing unit providing a plurality of labeling images according to labeling with a set color;
An identification object image providing unit which receives input from the user or an external identification object image requiring identification of a state including water;
Analyze an image feature using a plurality of patterned masks for a labeling image or the image to be labeled with at least one color, and obtain feature pattern components according to a feature item corresponding to each of the patterned masks A feature pattern component obtaining unit;
A feature map generator to create a feature map based on patterned masks used in the feature pattern component acquirer and feature pattern components obtained using the patterned masks; And
A comparison value calculator for comparing a feature pattern component of the identification target image with a feature map created by the feature map generator to calculate a comparison value;
Apparatus for determining the state of the water containing region of the video image using color labeling comprising a. The method of claim 7, wherein
A pattern mask generation unit generating a patterned mask by generating a mask having a predetermined size and giving a binarized value to each pixel constituting the mask to form a pattern;
Apparatus for determining the water containing area of the video image using color labeling further comprising. The method of claim 7, wherein
A feature item setting unit configured to set a feature item according to a shape of a pattern corresponding to each of the masks patterned according to a user input;
Apparatus for determining the water containing area of the video image using color labeling further comprising. The method of claim 7, wherein the feature pattern component acquisition unit,
Acquiring feature pattern components according to a result of a convolution operation on respective pixel values constituting the labeling image and pixel values constituting at least one patterned mask of the patterned masks;
An apparatus for judging the water containing area of a video image using phosphor color labeling. The method of claim 7, wherein the feature pattern component acquisition unit,
Image using the patterned masks for a first labeling image labeled with any one color and a second labeling image different from the first labeling image labeled with the same color used for labeling the first labeling image. Analyzing features and acquiring feature pattern components according to predetermined feature items corresponding to each of the patterned masks
An apparatus for judging the water containing area of a video image using phosphor color labeling. The method of claim 11, wherein the feature map creation unit,
A first value that is a result of a convolution operation of respective pixel values constituting the first labeling image and pixel values constituting at least one patterned mask of the patterned masks, and the second labeling Creating a feature map based on a second value obtained by multiplying a predetermined learning rate by a convolution operation of feature pattern components obtained from an image and pixel values constituting the patterned mask;
An apparatus for judging the water containing area of a video image using phosphor color labeling.

Images