CN116403009A - A Color Recognition Method Based on Template Matching - Google Patents

Abstract

The present invention relates to the technical field of image processing, in particular to a color recognition method based on template matching, including color template learning and color recognition, comprising the following steps: step (1) setting a rectangular region of interest; step (2) setting the region of interest Inner RGB image is converted into HSV space; step (3) calculates the hue and saturation histogram array; step (4) marks the color; step (5) calculates the hue and saturation histogram of the image to be identified; step (6) calculates the image to be identified and the distance from the template image; step (7) obtains the score; step (8) the template corresponding to the highest score is the color of the recognition image. This method is easy to operate, does not need to manually set too many parameters, automatically learns features, can flexibly expand the template library, and realize the purpose of recognizing multiple colors. s color.

Description

A Color Recognition Method Based on Template Matching

technical field

The invention relates to the technical field of image processing, in particular to a color recognition method based on template matching.

Background technique

Using color space to describe the color of objects, and to process and recognize colors is a very common application in the field of computer vision and image processing. Our commonly used RGB color space does not reflect the specific color information of the object very well, but the HSV space can decompose the expression of color into three factors: hue, saturation and brightness, which is very suitable for human understanding and more in line with The visual perception rules of the human brain express the lightness, hue, and vividness of colors very intuitively, which facilitates the comparison between colors. Therefore, converting the image from the RGB color space to the HSV color space for color recognition can more accurately identify the color of the object.

At present, color is one of the most intuitive attributes for people to perceive the world and describe objects. In the industry, there are a large number of scenes that need to distinguish products based on color, that is, given a picture of an object, the computer can determine what color people use to represent it. It, for example, divides clothes into red, yellow, blue and other categories and packs them separately. Usually, the process of distinguishing is done manually, but due to the biological characteristics of the human body, it is prone to misidentification. For enterprises, the cost of employment is also high. huge burden. .

Contents of the invention

The purpose of the present invention is to overcome the above-mentioned problems in the prior art and provide a color recognition method based on template matching to solve the problems of high artificial dependence, large external influence and low recognition accuracy in the traditional method.

Above-mentioned purpose is to realize through following technical scheme:

A color recognition method based on template matching, including color template learning and color recognition, comprising the following steps:

Step (1) sets a rectangular region of interest;

Step (2) converts the RGB image into the HSV space in the region of interest;

Step (3) calculates hue and saturation histogram array;

Step (4) marking color;

Step (5) calculate the hue and saturation histogram of the image to be identified;

Step (6) calculates the distance to be identified and the template image;

Step (7) obtains the score value;

The template corresponding to the highest score in step (8) is the color of the recognized image.

Further, step (1) is specifically setting a region of interest in the RGB image, the region of interest includes a recognition target, and setting a corresponding color array according to the color of the recognition target.

Further, the step (2) also includes obtaining the maximum value pmax and the minimum value pmin in the pixel r, g, and b components, wherein

pmax = max(r, g, b) (1)

pmin=min(r,g,b) (2).

Further, the step (3) is specifically: calculating the hue h, the formula is as follows:

Calculate the saturation s, the formula is as follows:

Generate hue and saturation histogram arrays arr_h and arr_s, the sizes are 360 and 256 respectively;

Among them, arr_h[i] (0≤i≤359) and arr_s[j] (0≤j≤255) respectively represent the number of pixels with hue value i and saturation value j in the current image.

Further, the step (5) is specifically to convert the image to be recognized into the HSV space according to the formulas (1)-(4), and obtain the hue and saturation histogram array of the image, denoted as cur_h[360] and cur_s [256].

Further, in the step (6), the Euclidean distance method is used to calculate the distance between the image to be identified and the template image, including:

Set the hue histogram of the template to be (0, 0, ..., n _i , ..., 0) in extreme cases, and the hue histogram to be recognized is (0, 0, .... .., n _j ,..., 0), each of the two histograms contains 360 data;

则计算待识别图与模板距离dist的公式如下：Among them, n _i =n _j =num, and i≠j, num represents the total number of pixels, and defines the maximum distance

The formula for calculating the distance dist between the image to be recognized and the template is as follows:

Further, in the step (6), the Manhattan distance method is used to calculate the distance to be identified and the template image, including:

Define the maximum distance max Dist=2×num, then the formula for calculating the distance dist between the image to be recognized and the template is as follows:

Further, the formula of the final score score in the step (7) is as follows:

Beneficial effect

A color recognition method based on template matching provided by the present invention is simple to operate, does not need to manually set too many parameters, automatically learns features, can flexibly expand the template library, realizes the purpose of recognizing multiple colors, and has high recognition accuracy, which can avoid Deviation from the original color of the object due to factors such as lighting, shadows, and reflections.

Description of drawings

Fig. 1 is a flow chart of a color recognition method based on template matching according to the present invention;

Fig. 2 is a template diagram of an embodiment in a color recognition method based on template matching according to the present invention;

Fig. 3 is a diagram of an image to be recognized in an embodiment of a color recognition method based on template matching according to the present invention.

Detailed ways

The present invention will be described in further detail below according to the drawings and embodiments. The described embodiments are only some, not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

As shown in Figure 1, a color recognition method based on template matching, including color template learning and color recognition,

The color template learning includes the following steps:

Step (1) sets a rectangular region of interest;

Step (2) converts the RGB image into the HSV space in the region of interest;

Step (3) calculates hue and saturation histogram array;

Step (4) marking color;

The color recognition includes the following steps:

Step (5) calculate the hue and saturation histogram of the image to be identified;

Step (6) calculates the distance to be identified and the template image;

Step (7) obtains the score value;

The template corresponding to the highest score in step (8) is the color of the recognized image.

Step (1) in this embodiment is specifically setting a region of interest in the RGB image, the region of interest includes a recognition target, and setting a corresponding color array according to the color of the recognition target. It also includes obtaining the maximum value pmax and minimum value pmin in the r, g, and b components of pixels, where:

pmax = max(r, g, b) (1)

pmin=min(r,g,b) (2).

The step (3) described in this embodiment is specifically: calculate the hue h, the formula is as follows:

Calculate the saturation s, the formula is as follows:

Generate hue and saturation histogram arrays arr_h and arr_s, the sizes are 360 and 256 respectively;

Among them, arr_h[i] (0≤i≤359) and arr_s[j] (0≤j≤255) respectively represent the number of pixels with hue value i and saturation value j in the current image.

The step (5) described in this embodiment is specifically: according to the formulas (1)-(4), the image to be recognized is converted into the HSV space, and the hue and saturation histogram array of the image is obtained, denoted as cur_h[360] and cur_s[256].

In step (6) described in the present embodiment, can adopt Euclidean distance method or Manhattan distance method to calculate the distance to be identified and template image, specifically as follows:

The Euclidean distance method is used to calculate the distance between the image to be recognized and the template image, including:

Set the hue histogram of the template to be (0, 0, ..., n _i , ..., 0) in extreme cases, and the hue histogram to be recognized is (0, 0, .... .., n _j ,..., 0), each of the two histograms contains 360 data;

则计算待识别图与模板距离dist的公式如下：Among them, n _i =n _j =num, and i≠j, num represents the total number of pixels, and defines the maximum distance

The formula for calculating the distance dist between the image to be recognized and the template is as follows:

The Manhattan distance method is used to calculate the distance between the image to be recognized and the template image, including:

Define the maximum distance max Dist=2×num, then the formula for calculating the distance dist between the image to be recognized and the template is as follows:

The formula for calculating the final score score in the step (7) is as follows:

There may be multiple templates, corresponding to multiple scores, and the bubble sort method is used to sort the scores from high to low. The higher the score, the greater the probability that the product to be recognized is the color.

This method mainly includes two steps of template learning (color template learning) and template matching (color recognition):

1) Template learning, under normal circumstances, the number of color types of a batch of products is known, and each product of each color is extracted, and an industrial camera is used to capture images, learn its color characteristics, and mark the specific color name;

2) Template matching, compare the similarity between the image of the product to be recognized and the template, and calculate several scores, the name of the template corresponding to the highest score is the actual color of the product to be recognized.

The colors involved in this proposal refer specifically to the visual effects produced by people through eye and brain activities, which are different from the definition of light wavelengths in the physical sense.

In order to verify the performance of the color recognition method based on template matching, this embodiment collects images of four color products, a total of 12 images are used for experimental testing, of which 4 are used as templates and 8 are used as test images.

First, template learning is carried out. There are four colors in the test image. Each image is selected as a template. As shown in Figure 2, rectangular regions of interest (need to include the product part in the image as much as possible) are set respectively, marked as red, green, blue and yellow;

Template matching is performed on 8 test images (as shown in Figure 3), and the similarity value is obtained by using the Manhattan distance method. The results are shown in Table 1.

Table 1 Sample matching similarity value and recognition results

Taking sample 2 in Table 1 as an example, the similarities between this sample and the four templates of red, green, blue, and yellow are 0.8775, 0.4973, 0.5093, and 0.5381, respectively, and the highest similarity with the red template is 0.8775, so the sample It is judged as red, and this sample corresponds to Figure 3-Sample 2. It can be seen from observation that the identification is correct.

The above description is only to illustrate the implementation of the present invention, and is not intended to limit the present invention. For those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention, All should be included within the protection scope of the present invention.

Claims (8)

1. a color recognition method based on template matching, is characterized in that, comprises color template study and color recognition, comprises the steps: Step (1) sets a rectangular region of interest; Step (2) converts the RGB image into the HSV space in the region of interest; Step (3) calculates hue and saturation histogram array; Step (4) marking color; Step (5) calculate the hue and saturation histogram of the image to be identified; Step (6) calculates the distance to be identified and the template image; Step (7) obtains the score value; The template corresponding to the highest score in step (8) is the color of the recognized image. 2. a kind of color recognition method based on template matching according to claim 1, is characterized in that, step (1) is specifically set interest area in described RGB image, and described interest area comprises recognition target, according to For the color of the identified target, set the corresponding color array. 3. a kind of color recognition method based on template matching according to claim 1, is characterized in that, described step (2) also comprises obtaining pixel r, g, maximum value pmax and minimum value pmin in component b, wherein pmax = max(r, g, b) (1) pmin=min(r,g,b) (2). 4. a kind of color recognition method based on template matching according to claim 3, is characterized in that, described step (3) is specifically: calculate hue h, and formula is as follows:

Calculate the saturation s, the formula is as follows:

Generate hue and saturation histogram arrays arr_h and arr_s, the sizes are 360 and 256 respectively; Among them, arr_h[i] (0≤i≤359) and arr_s[j] (0≤j≤255) respectively represent the number of pixels with hue value i and saturation value j in the current image. 5. a kind of color recognition method based on template matching according to claim 4, is characterized in that, described step (5) is specifically according to formula (1)～(4) image to be recognized is transformed into HSV space, Get the hue and saturation histogram array of the image, denoted as cur_h[360] and cur_s[256]. 6. a kind of color recognition method based on template matching according to claim 1, is characterized in that, adopts Euclidean distance method to calculate the distance to be identified and template image in the described step (6), comprising: Set the hue histogram of the template to be (0, 0, ..., n _i , ..., 0) in extreme cases, and the hue histogram to be recognized is (0, 0, .... .., n _j ,..., 0), each of the two histograms contains 360 data; Among them, n _i =n _j =num, and i≠j, num represents the total number of pixels, and defines the maximum distance

The formula for calculating the distance dist between the image to be recognized and the template is as follows:

7. a kind of color recognition method based on template matching according to claim 1, is characterized in that, adopts Manhattan distance method to calculate the distance to be identified and template image in the described step (6), comprising: Define the maximum distance maxDist=2×num, then the formula for calculating the distance dist between the image to be recognized and the template is as follows:

8. according to claim 6 or 7 described arbitrary a kind of color recognition method based on template matching, it is characterized in that, the formula of final score score in described step (7) is as follows:

Images