CN113591910A - Nixie tube display instrument identification method - Google Patents

Abstract

The invention discloses a method for identifying a nixie tube display instrument, which relates to the technical field of identifying a nixie tube and aims to realize intelligent identification of a nixie tube display type instrument, and comprises the following steps of: acquiring an RGB image, an instrument real-time image and a preset instrument image, and matching and transforming; cutting a real-time diagram of the instrument and a preset instrument diagram; performing K-Means clustering; assigning values to the instrument real-time graph; carrying out morphological transformation on the instrument real-time graph; extracting connected regions and subsets of the instrument real-time graph; identifying a single nixie tube number; splicing character strings; the string is converted to a floating point number. The intelligent high-accuracy nixie tube display instrument identification method is realized.

Description

Nixie tube display instrument identification method

Technical Field

The invention relates to the technical field of digital display tube identification, in particular to a method for identifying a digital display instrument.

Background

In industrial production today, more and more meters are installed and put into service in factories due to the automation of production and the more comprehensive monitoring of the production process. At present, reading and recording are mostly carried out on an instrument panel manually.

However, in many factories, the environment is severe, for example, radiation exists, and the project of reading data by entering the factory at regular time by workers is cumbersome and still may cause harm to human bodies, and meanwhile, the situation of too large error may occur in manual reading. The applications of the nixie tube display instrument are more and more extensive, and different from a pointer instrument, the identification of the nixie tube display instrument needs to judge aiming at the display of numbers.

The intelligent identification mode is used for replacing manual reading of the pointer instrument, so that the accuracy and the working efficiency can be improved.

Disclosure of Invention

The invention discloses a method for identifying a nixie tube display instrument, and aims to realize intelligent identification of a nixie tube display type instrument.

In order to achieve the purpose, the invention adopts the following technical scheme:

a nixie tube display instrument identification method comprises the following steps:

step 1: acquiring RGB (red, green and blue) graph and instrument real-time graph { I (real-time) of instrument scene_i,j} and obtaining a preset instrument map { I'_i,jMatching with preset image characteristic points to perform projection transformation, solving a perspective matrix A, and combining the { I }_i,jThe coordinates of the { i } are mapped to { i ″ }using the perspective matrix A_i,j}；

Step 2: according to the pre-transferred meter area information { x, y, w, h }, the result is passed { I ″ "_i,jCutting out, using the cutting out area as new { I }_i,jAre simultaneously paired with { I'_i,jClipping is carried out, and the clipping area is used as new { I'_i,j}；

And step 3: to { I_i,jPerforming K-Means clustering, wherein the number of clustering categories is 2;

and 4, step 4: according to each I_i,jClass of belonged and Preset Meter information, pair I_i,jAnd carrying out assignment according to the following assignment criteria:

and 5: to { I_i,jPerforming morphological transformation with I_i,jFor a center of 5X5 pixels, consider I_i,jPixel values in the neighborhood 5x5 range when there is a non-zero imageElement, then I_i,j＝255；

Step 6: extraction of { I_i,jCommunicated area of }

Take its subset S_i}，{S_iThe following conditions are satisfied:

(1)

S_j∈{S_j}，

(2)

and 7: identifying a single nixie tube number, for each S_iCarrying out identification;

and 8: sequentially splicing and identifying the obtained non-empty characters to a character string Str from left to right;

and step 9: and converting Str into floating point number as the identification result of the nixie tube display instrument.

Preferably, the step 1 comprises the steps of:

step 11: obtaining a real-time view of the meter { I_i,j}；

Step 12: obtain preset instrument diagram { I'_i,j}；

Step 13: respectively extracting real-time graphs { I of meters_i,jAnd preset gauge chart { I'_i,jExtracting main direction and then performing rotation invariance processing, and then extracting BRIEF feature descriptors;

step 14: comparing instrument real-time graphs { I }respectively_i,jAnd preset gauge chart { I'_i,jObtaining the minimum Hamming distance distH of each corresponding feature point descriptor in the Chinese character_min；

Step 15: extracting characteristic points p from the instrument real-time image and the preset instrument image respectively_iAnd q_iGroup (u) }Hamming distance less than 2 times distH_minI is the serial number of the characteristic point, and the serial numbers in p and q are consistent to form a characteristic point pair, wherein

Step 16: PA ≈ Q, solving perspective matrix

Wherein

Solving the optimal solution of A;

and step 17: will { I_i,jThe coordinates of the { are mapped to { I ″ } with A_i,j}。

Preferably, the step 16 finds the optimal solution by a least square method.

Preferably, said step 7 comprises the steps of:

(a) take its rectangular envelope region rect_iIts height and width are denoted row and col, respectively;

(b) if rect_iAn aspect ratio greater than a threshold of 5 is identified as "1";

(c) if rect_iPixel mean over range

Greater than a threshold value of 128, the character is recognized ".

(d) If rect_iPixel mean over range

Less than threshold 50, then empty is identified

(e) To pair

Traversing the row, recording the turnover times of the pixel values, and recording the turnover times ascM, i.e.:

(f) to pair

Is traversed, the number of flips is noted as cUL

(g) To pair

The pixel row of (2) is traversed, and the number of turns is recorded as cUR;

(h) to pair

Traversing the pixel rows, and recording the turnover times as cDL;

(i) to pair

The pixel row of (2) is traversed, and the number of turns is recorded as cDR;

(j) calculating the value of the code representing the numerical value:

code＝10000cM+1000cUL+100cUR+10cDL+cDR

(k) and judging the number according to the code value-number reference table.

The invention realizes the intelligent identification of the instrument by reading and mapping the image displayed by the nixie tube to carry out digital identification; mapping is carried out through a perspective matrix, wherein the perspective matrix adopts a least square method to obtain an optimal solution, so that the calculation and mapping precision is improved; the envelope and the code are taken one by one for single number, and the specific numerical value identification is carried out by referring to the comparison table, the identification accuracy is high, the burden of workers is lightened,

drawings

Fig. 1 is a schematic flow chart of a high-precision pointer instrument identification method in embodiment 1.

Detailed Description

Example 1

The invention discloses a method for identifying a nixie tube display instrument, which is shown in a flow chart of figure 1 and comprises the following steps:

step 1: acquiring RGB (red, green and blue) graph and instrument real-time graph { I (real-time) of instrument scene_i,j} and obtaining a preset instrument map { I'_i,jMatching with preset image characteristic points to perform projection transformation, solving a perspective matrix A, and combining the { I }_i,jThe coordinates of { I } are mapped to { I ″ } using the perspective matrix A_i,j}。

To implement step 1, the following steps are preferably taken:

step 11: obtaining a real-time view of the meter { I_i,j}；

Step 12: obtain preset instrument diagram { I'_i,j}；

Step 13: respectively extracting real-time graphs { I of meters_i,jAnd preset gauge chart { I'_i,jExtracting main direction and then performing rotation invariance processing, and then extracting BRIEF feature descriptors;

step 14: comparing instrument real-time graphs { I }respectively_i,jAnd preset gauge chart { I'_i,jObtaining the minimum Hamming distance distH of each corresponding feature point descriptor in the Chinese character_min；

Step 15: extracting characteristic points p from the instrument real-time image and the preset instrument image respectively_iAnd q_iThe Hamming distance is less than 2 times distH_minI is the serial number of the characteristic point, and the serial numbers in p and q are consistent to form a characteristic point pair, wherein

Step 16: PA ≈ Q, solving perspective matrix

Wherein

Solving the optimal solution of A; preferably, a least square method is used as a method for obtaining the optimal solution.

And step 17: will { I_i,jThe coordinates of the { are mapped to { I ″ } with A_i,j}。

And (5) entering the step 2 after the image acquisition and mapping are finished.

Step 2: according to the pre-transferred instrument area information { x, y, w, h }, passing

Cutting is carried out, and the cutting area is used as a new { I_i,jAre simultaneously paired with { I'_i,jClipping is carried out, and the clipping area is used as new { I'_i,jSpecifically, x and y in the instrument area information are respectively an x coordinate and a y coordinate of a breakpoint at the upper left corner of the cutting frame, and w and h are respectively the width and the height of the cutting frame;

and step 3: to { I_i,jPerforming K-Means clustering, wherein the number of clustering categories is 2;

and 4, step 4: according to each I_i,jClass of belonged and Preset Meter information, pair I_i,jAnd carrying out assignment according to the following assignment criteria:

and 5: to { I_i,jPerforming morphological transformation with I_i,jFor a center of 5X5 pixels, consider I_i,jPixel values in the neighborhood 5x5 range, when there are non-zero pixels, then I_i,j＝255；

Step 6: extraction of { I_i,jCommunicated area of }

Take its subset S_i}，{S_iThe following conditions are satisfied:

(1)

S_j∈{S_j}，

(2)

and 7: identifying a single nixie tube number, for each S_iCarrying out identification; in this embodiment, the following steps are respectively performed:

(a) take its rectangular envelope region rect_iThe height and width of the pattern are respectively marked as row and col, and the rectangular envelope is the minimum envelope capable of laying down the target pattern;

(b) if rect_iAn aspect ratio greater than a threshold of 5 is identified as "1";

(c) if rect_iPixel mean over range

Greater than a threshold value of 128, the character is recognized ".

(d) If rect_iPixel mean over range

Less than threshold 50, then empty is identified

(e) To pair

Traversing the columns, recording the turnover times of the pixel values, and recording the turnover times as cM, namely:

(f) to pair

Is traversed, the number of flips is noted as cUL

(g) To pair

The pixel row of (2) is traversed, and the number of turns is recorded as cUR;

(h) to pair

Traversing the pixel rows, and recording the turnover times as cDL;

(i) to pair

The pixel row of (2) is traversed, and the number of turns is recorded as cDR;

(j) calculating the value of the code representing the numerical value:

code＝10000cM+1000cUL+100cUR+10cDL+cDR

(k) the number is determined from a code value-number reference table as shown in the following table, particularly in the case where both codes 2020 and 202 correspond to the identification number 1:

code value-number reference table

And 8: sequentially splicing and identifying the obtained non-empty characters to a character string Str from left to right;

and step 9: and converting Str into floating point number as the identification result of the nixie tube display instrument.

Claims (4)

1. A nixie tube display instrument identification method is characterized by comprising the following steps:

step 1: acquiring RGB (red, green and blue) graph and instrument real-time graph { I (real-time) of instrument scene_i，j} and obtaining a preset instrument map { I'_i，jMatching with preset image characteristic points to perform projection transformation, solving a perspective matrix A, and combining the { I }_i，jThe coordinates of { I } are mapped to { I ″ } using the perspective matrix A_i，j}；

Step 2: according to the pre-transferred instrument area information { x, y, w, h }, the { I' is compared_i，jCutting out, using the cutting out area as new { I }_i，jAre simultaneously paired with { I'_i，jClipping is carried out, and the clipping area is used as new { I'_i，j}；

And step 3: to { I_i，jPerforming K-Means clustering, wherein the number of clustering categories is 2;

and 4, step 4: according to each I_i，jClass of belonged and Preset Meter information, pair I_i，jAnd carrying out assignment according to the following assignment criteria:

and 5: to { I_i，jPerforming morphological transformation with I_i，jFor a center of 5X5 pixels, consider I_i，jPixel values in the neighborhood 5x5 range, when there are non-zero pixels, then I_i，j＝255；

Step 6: extraction of { I_i，jCommunicated area of }

Take its subset S_i}，{S_iThe following conditions are satisfied:

(1)

(2)

and 7: identifying a single nixie tube number, for each S_iCarrying out identification;

and 8: sequentially splicing and identifying the obtained non-empty characters to a character string Str from left to right;

and step 9: and converting Str into floating point number as the identification result of the nixie tube display instrument.

2. The nixie tube display instrument recognition method as claimed in claim 1, wherein the step 1 comprises the steps of:

step 11: obtaining a real-time view of the meter { I_i，j}；

Step 12: obtain preset instrument diagram { I'_i，j}；

Step 13: respectively extracting real-time graphs { I of meters_i，jAnd preset gauge chart { I'_i，jExtracting main direction and then performing rotation invariance processing, and then extracting BRIEF feature descriptors;

step 14: comparing instrument real-time graphs { I }respectively_i，jAnd preset gauge chart { I'_i，jObtaining the minimum Hamming distance distH of each corresponding feature point descriptor in the Chinese character_min；

Step 15: extracting characteristic points p from the instrument real-time image and the preset instrument image respectively_iAnd q_iThe Hamming distance is less than 2 times distH_minI is the serial number of the characteristic point, and the serial numbers in p and q are consistent to form a characteristic point pair, wherein

Step 16: PA ≈ Q, solving perspective matrix

Wherein

Solving the optimal solution of A;

and step 17: will { I_i，jThe coordinates of the { are mapped to { I ″ } with A_i，j}。

3. The method as claimed in claim 2, wherein said step 16 is performed by a least square method to obtain the optimal solution.

4. The nixie tube display instrument recognition method as recited in claim 1, wherein the step 7 comprises the steps of:

(a) take its rectangular envelope region rect_iTheir height and width are denoted row and co/;

(b) if rect_iAn aspect ratio greater than a threshold of 5 is identified as "1";

(c) if rect_iPixel mean over range

Greater than a threshold value of 128, the character is recognized ".

(d) If rect_iPixel mean over range

Less than threshold 50, then empty is identified

(e) To pair

Traversing the columns, recording the turnover times of the pixel values, and recording the turnover times as cM, namely:

(f) to pair

Is traversed, the number of flips is noted as cUL

(g) To pair

The pixel row of (2) is traversed, and the number of turns is recorded as cUR;

(h) to pair

Traversing the pixel rows, and recording the turnover times as cDL;

(i) to pair

The pixel row of (2) is traversed, and the number of turns is recorded as cDR;

(j) calculating the value of the code representing the numerical value:

code＝10000cM+1000cUL+100cUR+10cDL+cDR

(k) and judging the number according to the code value-number reference table.

Images