Multi-source optical sensor information fusion color identification method based on evidence theory

Abstract

The invention discloses an evidence theory-based color identification method for information fusion of a multi-source optical sensor. The method comprises the following steps of: acquiring RGB (Red, Green and Blue) numerical values of different photoelectric sensors; respectively constructing triangular fuzzy numbers based on the acquired RGB data; constructing a plurality of combined triangular fuzzy numbers for different colors; carrying out membership distribution; basic probability distribution functions of the sensors under different color attributes are generated based on the values of the membership functions; fusing different color data information of the three sensors; and fusing the BPA of the three sensors again to generate a color identification result. According to the invention, the color identification precision is improved by fusing the signal results of the plurality of light sensors; and a plurality of light sensors are integrated for color recognition, so that the stability and robustness of color recognition are improved.

Description

Technical Field

[0001] The invention relates to color recognition of an object, and in particular to a color recognition method based on multi-source light sensor information fusion based on evidence theory. Background Art

[0002] Color recognition of objects is a widely needed application technology, which is widely used in scientific research, art design, decoration, film and television game design and industrial production. The existing color recognition principle based on the light spot effect is based on the physical property that different structures on the surface of an object will reflect different light. When the light source shines on the surface of an object, the object will reflect light of a specific wavelength for a specific color characteristic according to its own structure and material characteristics. These lights contain the surface structure and color information of the object. These reflected lights are received by the photosensitive elements based on the photoelectric effect in the color recognition sensor, and the light signals can be converted into electrical signals. Subsequently, the analog or digital circuits of the signal processing amplify and filter these electrical signals, and then formulate specific color recognition algorithms (such as RGB algorithms, HSV algorithms, etc.) based on different standards to identify the color of the object.

[0003] From the above color recognition principle, we can see that color recognition is actually a specific classification based on the characteristics of the light reflected by the surface material structure of the object to be recognized. In daily use, most detections use a single light sensor for color recognition, such as Figure 2 As shown, this single sensor performs color recognition, and the recognition stability is poor and the relative accuracy is low. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide a color recognition method based on multi-source light sensor information fusion based on evidence theory.

[0005] Technical solution: The color recognition method of multi-source light sensor information fusion based on evidence theory of the present invention comprises the following steps:

[0006] (1) Collecting RGB values ​​of different photoelectric sensors, that is, triangular fuzzy numbers based on the photoelectric sensor data of R, G, and B elements. The RGB value range collected by a single sensor is a set of 3 data;

[0007] (2) construct triangular fuzzy numbers based on the collected RGB data;

[0008] (3) Construct multiple combined triangular fuzzy numbers for different colors;

[0009] (4) The values ​​of RGB color attributes obtained by different photoelectric sensors are respectively placed into the three different triangular fuzzy number sets mentioned above to perform membership assignment;

[0010] (5) generating basic probability distribution functions of each sensor under different color attributes based on the values ​​of the above-mentioned membership functions;

[0011] (6) Fusing the different color data from the three sensors;

[0012] (7) The BPA, or basic probability allocation function, of the three sensors is fused again to generate the color recognition result.

[0013] Furthermore, the step (1) selects an RGB color standard based on the industry, that is, a color benchmark is determined by changes in the three color channels of red, green and blue (RGB).

[0014] Furthermore, the photoelectric sensor in step (1) is an RGB sensor, which is a sensor that obtains different values ​​of R, G, and B by processing the reflected light from the surface of the object through the photoelectric effect.

[0015] Furthermore, in step (2), triangular fuzzy numbers are constructed for the collected RGB data, respectively, which are expressed as:

[0016]

[0017] in, It represents the degree of membership of the collected RGB data for different colors. x is the value of a single R, G or B. a represents the minimum value of a certain type of color. b represents the average single color value of a certain type of color. c represents the maximum value of a certain type of color.

[0018] Furthermore, in the step (3), the RGB three-primary color light mode is selected, and the test value of the red attribute is 127, the test value of the green attribute is 255, and the test value of the blue attribute is 212.

[0019] Furthermore, the degree of membership in step (4) is represented by the height of a new triangle formed by the intersection of two triangles, and the formula used is as follows:

[0020]

[0021] Among them, μ represents the degree of membership between the object to be tested and the standard object proposition; k represents the value measured by the sensor; 1 represents the difference between the test value and the maximum or minimum value when the test value of the sensor is in the range between the maximum and minimum values ​​of the standard object; d represents the difference between the average value of the standard sample and its maximum or minimum value when the test value of the sensor is in the range between the maximum and minimum values ​​of the standard object; A represents the proposition to be compared with the test result.

[0022] Furthermore, the step (4) uses different RGB values ​​obtained by three sensors for display, and the values ​​are: sensor 1: {116, 242, 200}; sensor 2: {117, 245, 203}; sensor 3: {119, 244, 201}.

[0023] Furthermore, the basic probability distribution function of each sensor under different color attributes generated in step (5) is expressed as:

[0024]

[0025] Among them, p satisfies: 1≤p≤n, and the number of focal elements generated at last is p+1; Indicates the membership degree of a proposition, and is replaced by μ in the test data of the experimental part; It represents the unknown part of the generated BPA, and its specific expression is: Indicates that the test data supports the proposition T i degree of affiliation;

[0026] T j represents a proposition; Ω represents the universe, which is the exhaustive set of all possible values ​​of variable B; m(T j ) represents the proposition T j The generated basic probability distribution function; m(Ω) represents the uncertainty of the type of object to be tested in the generation of the basic probability distribution function.

[0027] Furthermore, the information fusion formula used in step (6) is:

[0028]

[0029] Among them, K is the conflict coefficient, which indicates the conflict between two pieces of evidence, A i represents the i-th color proposition, B j represents the jth color proposition, m1 represents the sensor 1 to A i The basic probability assignment function of the color proposition, m2 represents the second sensor pair B j The basic probability assignment function for a color proposition.

[0030] Furthermore, the conflict coefficient K is calculated as follows:

[0031]

[0032] Among them, A i Indicates that B j represents, m1 represents, m2 represents.

[0033] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: based on the evidence theory of information fusion, the accuracy of color recognition is improved by fusing the signal results of multiple optical sensors; multiple optical sensors are integrated for color recognition, which improves the stability and robustness of color recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flowchart of information fusion of multiple sensors;

[0035] Figure 2 Flowchart for a single sensor;

[0036] Figure 3 Schematic diagram of the constructed triangular fuzzy number;

[0037] Figure 4 Schematic diagram of triangular fuzzy numbers with red attributes;

[0038] Figure 5 Schematic diagram of triangular fuzzy numbers with green attributes;

[0039] Figure 6 Schematic diagram of triangular fuzzy numbers with blue attributes;

[0040] Figure 7 It is a schematic diagram of the membership distribution of the red attribute R in triangular fuzzy numbers;

[0041] Figure 8 It is a schematic diagram of the membership distribution of the green attribute R in triangular fuzzy numbers;

[0042] Fig. 9 Schematic diagram of the membership distribution of the blue attribute R in triangular fuzzy numbers. DETAILED DESCRIPTION

[0043] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0044] like Figure 1 As shown, the color recognition method of multi-source light sensor information fusion based on evidence theory of the present invention includes:

[0045] (1) Collect RGB values ​​from different photoelectric sensors. Traditionally, the range of a single RGB value is 0-255. For example, the RGB value of red is (255, 0, 0). Therefore, the RGB value range collected by a single sensor is a set of 3 data;

[0046] First, it is necessary to select a color classification standard. The present invention is based on the RGB color standard in the industry, that is, the color is determined by the changes in the three color channels of red, green and blue (RGB). This standard covers all colors that human vision can perceive. Therefore, the optical sensor based on the present invention is an RGB sensor, that is, a sensor that obtains different values ​​of R, G, and B after processing the reflected light from the surface of an object through the photoelectric effect.

[0047] Then the R, G, and B values ​​obtained by the sensor are placed in the corresponding classification ranges of different color categories, and the evidence recognition framework of the evidence theory is constructed according to the classification range.

[0048] A triangular fuzzy tree is constructed based on the recognition framework and the values ​​of each element (RGB).

[0049] Based on the triangular fuzzy values ​​of the photoelectric sensor data of the R, G, and B elements, the corresponding basic probability assignment function (BPA) is generated.

[0050] Based on the BPA credibility fusion rule of evidence theory, the BPAs generated by different light sensors are fused, the fused BPAs are converted into probabilities, and decisions are made on the colors of detected objects.

[0051] (2) construct triangular fuzzy numbers based on the collected RGB data;

[0052] The following formula is used to construct the triangular fuzzy number of a single RGB value:

[0053]

[0054] in, It represents the degree of membership of the collected RGB data for different colors. x is the value of a single R, G or B. a represents the minimum value of a certain type of color. b represents the average single color value of a certain type of color. c represents the maximum value of a certain type of color.

[0055] The constructed triangular fuzzy number is as follows Figure 3 As shown:

[0056] (3) Construct multiple combined triangular fuzzy numbers for different colors; construct multiple combined triangular fuzzy numbers for different colors. For example, in order to identify Aquamarine, a relatively common cyan-green color, in the RGB three-primary color light mode, the test value of its red attribute is generally 127, the test value of its green attribute is generally 255; the test value of its blue attribute is generally 212. The wavelength of Aquamarine is about 505.78 nanometers. Aquamarine 1, 2, and 3 are three color categories in this color system. Because the proportions of the three primary colors of the three colors are very close, it is difficult to distinguish them manually. Therefore, in daily industrial production, we often use photoelectric color recognition sensors to identify and distinguish colors. The comparison table of the test values ​​of the three primary colors R, G, and B for these three colors is as follows:

[0057] Table 1: Comparison of the test values ​​of the three primary colors R, G, and B for three colors

[0058]

[0059] Based on this, we can construct the triangular fuzzy number sets of the three sub-color systems respectively. The triangular fuzzy number of the red attribute is as follows: Figure 4 As shown: The triangular fuzzy number of the green attribute is Figure 5 As shown: The triangular fuzzy number of the blue attribute is Figure 6 As shown: A1 represents Aquamarine 1, A2 represents Aquamarine 2, and A3 represents Aquamarine 3. It can be clearly seen that these three colors have a lot of overlap in red and blue attributes, which is why it is very difficult for the human eye to identify these colors.

[0060] (4) The values ​​of RGB color attributes obtained by different photoelectric sensors are placed in the three different triangular fuzzy number sets mentioned above to assign membership. As an important parameter for generating BPA in this paper, membership can be expressed by the height of the new triangle formed by the intersection of two triangles. The formula required is as follows:

[0061]

[0062] Among them, μ represents the degree of membership between the object to be tested and the standard object proposition; k represents the value measured by the sensor; 1 represents the difference between the test value and the maximum or minimum value when the test value of the sensor is between the maximum and minimum values ​​of the standard object; d represents the difference between the average value of the standard sample and its maximum or minimum value when the test value of the sensor is between the maximum and minimum values ​​of the standard object; A represents the proposition compared with the test result. Different RGB values ​​obtained by three sensors are used for display here, and their values ​​are: Sensor 1: {116, 242, 200}; Sensor 2: {117, 245, 203}; Sensor 3: {119, 244, 201}. (All are displayed in the order of RGB.) The membership distribution of the red attribute R obtained by the three sensors in the triangular fuzzy number is shown as follows Figure 7 As shown; its membership distribution in the triangular fuzzy number of the green attribute G is as follows Figure 8 As shown; the membership distribution of the triangular fuzzy number of the blue attribute B is as follows Fig. 9 As shown in the figure, the three sensors belong to different color categories in three different color attributes, where A1 represents Aquamarine 1, A2 represents Aquamarine 2, and A3 represents Aquamarine 3. In triangular fuzzy numbers, the overlapping part is where the evidence theory performs the credibility distribution.

[0063] (5) Based on the values ​​of the above-mentioned membership functions, the basic probability distribution functions of each sensor under different color attributes are generated respectively; the generated mathematical formulas are as follows: The BPA calculation formulas of each proposition are as follows:

[0064]

[0065] Among them, p satisfies: 1≤p≤n, and the number of focal elements generated at last is p+1; Indicates the membership degree of a proposition, and is replaced by μ in the test data of the experimental part; It represents the unknown part of the generated BPA, and its specific expression is: Indicates that the test data supports the proposition T i degree of affiliation;

[0066] T j represents a proposition; Ω represents the universe, which is the exhaustive set of all possible values ​​of variable B; m(T j ) represents the proposition T jThe generated basic probability distribution function; m(Ω) represents the uncertainty of the type of object to be measured in the generation of the basic probability distribution function. The BPA generation results of the three sensors under different color attributes are shown in Table 2, Table 3, and Table 4:

[0067] Table 2: BPA of sensor 1 for the three attributes of R, G, and B

[0068]

[0069] Table 3: BPA of sensor 2 for the three attributes R, G, and B

[0070]

[0071] Table 4: BPA of sensor 3 for the three attributes R, G, and B

[0072]

[0073]

[0074] So far, the BPA generation of three sensors under three different color attributes has been completed.

[0075] (6) Fusing the different color data of the three sensors; this step mainly uses the following information fusion formula:

[0076]

[0077] Among them, K is the conflict coefficient, which indicates the conflict between two pieces of evidence, A i represents the i-th color proposition, B j represents the jth color proposition, m1 represents the sensor 1 to A i The basic probability assignment function of the color proposition, m2 represents the second sensor pair B j The basic probability assignment function for a color proposition.

[0078] Furthermore, the conflict coefficient K is calculated as follows:

[0079]

[0080] Among them, A i Indicates that B j represents, m1 represents, m2 represents.

[0081] The K obtained by the different color attributes of the three sensors is 1. After fusion, the BPA of the three sensors for color recognition is shown in Table 5, Table 6, and Table 7:

[0082] Table 5: BPA after sensor 1 evidence theory fusion

[0083]

[0084] Table 6: BPA after sensor 2 evidence theory fusion

[0085]

[0086] Table 7: BPA after theoretical fusion of sensor 3 evidence

[0087]

[0088] The table above shows the confidence distribution of the physical surface color classification obtained by the three sensors.

[0089] (7) The BPAs (basic probability allocation functions) of the three sensors are fused again to generate the color recognition results, as shown in Table 8:

[0090] Table 8: BPA of three sensors after evidence theory fusion

[0091]

[0092] It is obvious that the color detected by all three sensors is A2, which is Aquamarine 2.

Claims

1. A color recognition method based on multi-source light sensor information fusion based on evidence theory, characterized in that: The steps include: (1) Collecting RGB values ​​of different photoelectric sensors, that is, triangular fuzzy numbers based on the photoelectric sensor data of R, G, and B elements. The RGB value range collected by a single sensor is a set of 3 data; (2) construct triangular fuzzy numbers based on the collected RGB data; (3) Construct multiple combined triangular fuzzy numbers for different colors; (4) The values ​​of RGB color attributes obtained by different photoelectric sensors are respectively placed into the three different triangular fuzzy number sets mentioned above to perform membership assignment; (5) generating basic probability distribution functions of each sensor under different color attributes based on the values ​​of the above-mentioned membership functions; (6) Fusing the different color data from the three sensors; (7) The BPA, or basic probability allocation function, of the three sensors is fused again to generate the color recognition result.

2. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The step (1) selects the RGB color standard based on the industry, that is, the color benchmark is determined by the changes in the three color channels of red, green and blue (RGB).

3. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The photoelectric sensor in step (1) is an RGB sensor, which is a sensor with different R, G, and B values ​​obtained by processing the reflected light from the surface of an object through the photoelectric effect.

4. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The step (2) constructs triangular fuzzy numbers for the collected RGB data, which can be expressed as: in, It represents the degree of membership of the collected RGB data for different colors. x is the value of a single R, G or B. a represents the minimum value of a certain type of color. v represents the average single color value of a certain type of color. c represents the maximum value of a certain type of color.

5. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: In the step (3), the RGB three-primary color light mode is selected, and the test value of the red attribute is 127, the test value of the green attribute is 255, and the test value of the blue attribute is 212.

6. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The degree of membership in step (4) is represented by the height of a new triangle formed by the intersection of two triangles. The formula used is as follows: Among them, μ represents the degree of membership between the object to be tested and the standard object proposition; k represents the value measured by the sensor; 1 represents the difference between the test value and the maximum or minimum value when the test value of the sensor is in the range between the maximum and minimum values ​​of the standard object; d represents the difference between the average value of the standard sample and its maximum or minimum value when the test value of the sensor is in the range between the maximum and minimum values ​​of the standard object; A represents the proposition to be compared with the test result.

7. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The step (4) uses different RGB values ​​obtained by three sensors for display, and the values ​​are: sensor 1: {116, 242, 200}; sensor 2: {117, 245, 203}; sensor 3: {119, 244, 201}.

8. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The basic probability distribution function generated by step (5) for each sensor under different color attributes is expressed as: Among them, p satisfies: 1≤p≤n, and the number of focal elements generated at last is p+1; Indicates the membership degree of a proposition, and is replaced by μ in the test data of the experimental part; It represents the unknown part of the generated BPA, and its specific expression is: Indicates that the test data supports the proposition T i degree of affiliation; T j represents a proposition; Ω represents the universe, which is the exhaustive set of all possible values ​​of variable B; m(T j ) represents the proposition T j The generated basic probability distribution function; m(Ω) represents the uncertainty of the type of object to be tested in the generation of the basic probability distribution function.

9. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 1 is characterized in that: The information fusion formula used in step (6) is: Among them, K is the conflict coefficient, which indicates the conflict between two pieces of evidence, A i represents the i-th color proposition, B j represents the jth color proposition, m1 represents the sensor 1 to A i The basic probability assignment function of the color proposition, m2 represents the second sensor pair B j The basic probability assignment function for a color proposition.

10. The color recognition method based on multi-source light sensor information fusion based on evidence theory according to claim 9 is characterized in that: The conflict coefficient K is calculated as follows: Among them, A i represents the i-th color proposition, B j represents the jth color proposition, m1 represents the sensor 1 to A i The basic probability assignment function of the color proposition, m2 represents the second sensor pair B j The basic probability assignment function for a color proposition.

Images