Multi-class mechanical water treatment metering instrument identification and reading method

Abstract

The invention discloses a multi-class mechanical water treatment metering instrument reading method based on machine vision, and the method comprises the steps: collecting and preprocessing a video, building an instrument image database, building an instrument type recognition algorithm, and carrying out the recognition of the type of an instrument which is recognized as a pointer type instrument. The position of a pointer tip relative to two nearest scale numbers is obtained through instrument scale number recognition and pointer position recognition of the pointer type instrument, and then instrument reading is obtained through calculation after the distance proportion between the point of the pointer and the two nearest numbers is calculated. For the instrument identified as the floater type instrument, the position of the upper edge of the floater and two scale readings closest to the position are obtained through instrument scale number identification and floater position identification, and then the numerical value corresponding to the coordinate where the intersection point of the center connecting line of the two scale numbers and the upper edge of the floater is located is calculated. According to the method, the aim of automatically identifying various common mechanical instruments of the water treatment device can be fulfilled by fusing machine vision with machine learning, and the method has the advantages of high real-time performance, high universality and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of digital image processing, and particularly to a method for identifying and reading multi-category mechanical water treatment metering instruments. Background Art

[0002] In water treatment equipment, instruments are usually divided into analytical instruments for measuring water quality and metering instruments for measuring water pressure, water temperature, flow rate, liquid level height, etc. In most cases, in order to improve the automation and remote monitoring capabilities of the equipment, two sets of instruments, namely local mechanical metering instruments and remote analog instruments, are configured. That is, the local instruments are used for daily inspection records, and the remote instruments are used for the central control system to read analog data for automatic recording and monitoring.

[0003] Patent CN112488030A proposes a machine vision recognition method for pointer-type instruments, and patent CN115273093A proposes a machine vision recognition method for rotary-wheel type instruments. In the existing patents, most are recognition methods for pointer-type instruments, and the recognition schemes designed by the methods mostly use feature detection and feature point matching methods for recognition. The applicability flexibility of the instruments is not high, and there are problems such as being unable to recognize other instrument types and instrument panels outside the prefabricated instruments.

[0004] In addition, the common metering instrument types in water treatment equipment can be divided into three categories: pointer-type instruments for measuring water pressure, water temperature, etc., float-type instruments for measuring flow rate, and liquid surface-type instruments for measuring liquid level height. The types of instruments are more complex, and the existing patents have not solved the problem of automatic recognition of multi-type instruments.

[0005] In view of this, there is a large room for improvement in the existing technology. Summary of the Invention

[0006] The purpose of the present invention is to solve the above problems and design a method for identifying and reading multi-category mechanical water treatment metering instruments.

[0007] To achieve the above purpose, the technical solution of the present invention is a method for identifying and reading multi-category mechanical water treatment metering instruments, including the following steps:

[0008] Step 1: The video data collected by the camera is preprocessed through grayscale conversion and binarization.

[0009] Step 2: The instrument type is obtained by matching through a machine learning image recognition algorithm in the instrument image database.

[0010] Step 3.1: If it is identified as a pointer-type instrument, after two steps of identifying the instrument scale numbers and the pointer position, finally, the instrument reading is obtained by the proportional method.

[0011] Step 3.2: If it is identified as a float-type instrument, after two steps of instrument scale number recognition and float position recognition, finally, the instrument reading is obtained by the intersection method;

[0012] Step 3.3: If it is identified as a liquid-level instrument, it is necessary to further determine whether it is an external liquid-level instrument or an immersed liquid-level instrument in the instrument image database through a machine learning image recognition algorithm. After two steps of background scale number recognition and liquid-level position recognition, if it is an external instrument, the instrument reading is obtained by the same intersection method as the float-type instrument; if it is an immersed instrument, the reading is obtained by the scale number reasoning method.

[0013] As a further description of this technical solution, in step 1, common image preprocessing methods including grayscale conversion, denoising, binarization, edge detection, and morphological processing can be used in the preprocessing process.

[0014] As a further description of this technical solution, in step 2, the instrument image database is a pre-established database that can provide training for machine learning algorithms, and the database stores picture and video data of various pointer-type instruments, float-type instruments, and liquid-level instruments.

[0015] As a further description of this technical solution, in step 2, the machine learning image recognition algorithm uses a combined algorithm form including artificial neural networks, support vector machines, and various existing algorithms.

[0016] As a further description of this technical solution, in step 3.1, step 3.2, and step 3.3, the scale number recognition is obtained through a convolutional neural network algorithm, and the digital edge range is obtained, and further the digital center point coordinates are obtained, providing a basis for subsequent reading.

[0017] As a further description of this technical solution, in step 3.1, step 3.2, and step 3.3, the pointer position recognition, float position recognition, and liquid-level position recognition are obtained through edge recognition algorithms such as Roberts operator, Prewitt operator, Sobel operator, Canny operator, and Laplacian operator, and finally the coordinate data of the pointer tip, the upper edge of the float, and the position of the liquid level are obtained, providing a basis for subsequent reading.

[0018] As a further description of this technical solution, the specific steps for reading the instrument reading by the proportional method in step 3.1 are: (1) the coordinates of the center points of each digit and the pointer tip coordinates; (2) calculate the distance ratio x:y between the tip and the two closest digits (a and b); (3) obtain the final reading through the calculation formula (1 - x)a / y + xb / y.

[0019] As a further description of the present technical solution, in step 3.2 and step 3.3, the specific steps for reading the instrument reading by the intersection method are as follows: (1) The coordinates of the center points of each digit and the coordinates of the position where the upper edge of the float is located or the coordinates of the interface line between the liquid surface and the air; (2) Calculate the coordinates of the point where the line connecting the center points of two scale digits intersects the upper edge of the float (or the interface line between the liquid surface and the air) or its extension line; (3) Calculate the distance ratio x:y between the intersection point and the nearest two digits (a and b); (3) Obtain the final reading through the calculation formula (1 - x)a / y + xb / y.

[0020] As a further description of the present technical solution, the specific steps for reading the instrument reading by the scale digit reasoning method in step 3.3 are as follows: (1) According to the coordinates of the center points of each digit and the coordinates of the interface line between the liquid surface and the air; (2) Obtain the coordinates of the intersection point of the line connecting the midpoint of the digits and the liquid surface line; (3) Obtain the two digits closest to the intersection point (a and b from near to far); (4) Calculate the distances between the center points of the two digits and the distance from the intersection point to the center point of the nearest digit (x and y respectively); (5) Obtain the final reading through the calculation formula a - (b - a)y / x.

[0021] The beneficial effects are as follows. The present invention can automatically identify multiple types of instruments according to the instrument image data stored in the data through machine learning methods, and at the same time overcome the shortcoming that non-precast instrument panels in the prior art cannot be effectively identified; at the same time, through machine vision combined with machine learning algorithms such as convolutional neural networks, the reading recognition of pointer-type instruments, float-type instruments, and liquid surface-type instruments is carried out. Through the integration of machine vision and machine learning, the purpose of automatically identifying common mechanical instruments in multiple water treatment devices is achieved, and it has the advantages of high real-time performance, strong versatility, and low hardware cost compared with traditional analog monitoring instrument devices. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is the flow chart of the present invention;

[0023] Figure 2 is a schematic diagram of examples of common water treatment mechanical metering instruments covered by the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0024] The present invention will be specifically described below with reference to the accompanying drawings. As Figure 1 - Figure 2 shown, a method for identifying and reading multi-category mechanical water treatment metering instruments includes the following steps:

[0025] Step 1: The video data collected by the camera collects image data once per second. After preprocessing such as grayscale conversion, denoising, binarization, edge detection, and morphological processing, it enters the image recognition algorithm process.

[0026] Step 2: Match the instrument type in the instrument image database through a machine learning image recognition algorithm;

[0027] The image recognition algorithm is a classifier, and the classifier uses a combined algorithm form of multiple existing algorithms such as artificial neural networks and support vector machines.

[0028] The instrument image database is a pre-established database that can provide training for machine learning algorithms. The database stores picture and video data of various pointer-type instruments, float-type instruments, and liquid-level instruments (including external and immersion types).

[0029] Step 3.1: If it is recognized as a pointer-type instrument, such as Figure 2 where a is a mechanical pressure gauge in it. After two steps of identifying the instrument scale numbers and the pointer position, finally read the instrument reading through the ratio method.

[0030] The specific steps for reading the instrument reading of a pointer-type pressure gauge through the ratio method are: (1) Identify the coordinates of the center points of each recognized number and the coordinates of the pointer tip; (2) Calculate the distance ratio x:y between the tip and the two nearest numbers (a and b); (3) Obtain the final reading through the calculation formula (1 - x)a / y + xb / y.

[0031] Step 3.2: If it is recognized as a float-type instrument, such as Figure 2 where b is a float flowmeter in it. After two steps of identifying the instrument scale numbers and the float position, finally read the instrument reading through the intersection method.

[0032] The specific steps for reading the instrument reading of a float flowmeter through the intersection method are: (1) Identify the coordinates of the center points of each number and the coordinates of the position where the upper edge of the recognized float is located; (2) Calculate the coordinates of the point where the line connecting the centers of two scale numbers intersects with the upper edge of the float or its extension line; (3) Calculate the distance ratio x:y between the intersection point and the two nearest numbers (a and b); (3) Obtain the final reading through the calculation formula (1 - x)a / y + xb / y.

[0033] Step 3.3: If it is recognized as a liquid-level instrument, such as Figure 2 where c is a transparent tube liquid level gauge in it. After two steps of identifying the background scale numbers and the liquid level position, if it is an external liquid-level instrument, read the instrument reading through the same intersection method as the float-type instrument; if it is an immersion-type instrument, such as Figure 2 the liquid level scale in d, also after two steps of identifying the background scale numbers and the liquid level position, finally read the reading through the scale number reasoning method.

[0034] Among them, such as Figure 2For the transparent tube liquid level gauge of c, the specific steps for obtaining the instrument reading by the intersection method of this type of external liquid level instrument are as follows: (1) Identify the coordinates of the center points of each digit and the coordinates of the intersection line between the liquid level and the air obtained; (2) Calculate the coordinates of the point where the line connecting the centers of two scale digits intersects the intersection line between the liquid level and the air or its extension line; (3) Calculate the distance ratio x:y between the intersection point and the two nearest digits (a and b); (3) Obtain the final reading through the calculation formula (1 - x)a / y + xb / y.

[0035] For the immersion liquid level instrument, such as Figure 2 where d is the liquid level scale in c, the specific steps for reading the instrument reading by the reasoning method are as follows: (1) Identify the coordinates of the center points of each digit obtained and the coordinates of the intersection line between the liquid level and the air obtained; (2) Obtain the coordinates of the intersection point where the line connecting the midpoint of the digits intersects the line where the liquid level is located; (3) Obtain the two digits closest to the intersection point (a and b from near to far); (4) Obtain the distances between the centers of the two digits and between the intersection point and the center point of the nearest digit (x and y respectively); (5) Obtain the final reading through the calculation formula a - (b - a)y / x.

[0036] In addition, it should be noted that:

[0037] The scale digit recognition described in steps 3.1, 3.2, and 3.3 is obtained through the convolutional neural network algorithm, and the digital edge range is obtained, and further the coordinates of the digital center points are obtained, providing a basis for subsequent readings.

[0038] The pointer position recognition, float position recognition, and liquid level position recognition described in steps 3.1, 3.2, and 3.3 are obtained through edge recognition algorithms such as Roberts operator, Prewitt operator, Sobel operator, Canny operator, and Laplacian operator. Finally, the coordinate data of the pointer tip, the upper edge of the float, and the position where the liquid level is located are obtained, providing a basis for subsequent readings.

[0039] Among them, machine learning, artificial neural network, support vector machine, convolutional neural network, and edge recognition algorithm are prior arts.

[0040] The present invention can automatically identify various instrument types according to the instrument image data stored in the data through machine learning methods, and at the same time overcome the shortcoming that non-precast instrument panels in the prior art cannot be effectively recognized; at the same time, through machine vision combined with machine learning algorithms such as convolutional neural network, the readings of pointer-type instruments, float-type instruments, and liquid level-type instruments are recognized. Through the integration of machine vision and machine learning, the purpose of automatically identifying various mechanical instruments commonly used in water treatment devices is achieved, and it has the advantages of high real-time performance, strong versatility, and lower hardware costs compared with traditional analog monitoring instrument devices.

[0041] The above technical solutions only reflect the preferred technical solutions of the technical solutions of the present invention. Some changes that those skilled in the art of this technology may make to some parts thereof all reflect the principles of the present invention and fall within the protection scope of the present invention.

Claims

1. A method for identifying and reading multiple types of mechanical water treatment metering instruments, characterized in that: The following steps are involved: Step 1: The video data collected by the camera is preprocessed by grayscale and binarization; Step 2: Use machine learning image recognition algorithms to match instrument types in the instrument image database; Step 3.1: If it is a pointer instrument, after two steps of instrument scale number identification and pointer position identification, the instrument reading is finally read by the proportional method; Step 3.2: If it is identified as a float type instrument, after two steps of instrument scale digital identification and float position identification, the instrument reading is finally read by the intersection method; Step 3.3: If it is identified as a liquid level meter, it is necessary to further use the machine learning image recognition algorithm to determine whether it is an external liquid level meter or an immersed liquid level meter in the meter image database. After the two steps of background scale digital recognition and liquid level position recognition, if it is an external meter, the meter reading is read using the same intersection method as the float meter. If it is an immersed meter, the meter reading is read using the scale digital inference method.

2. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 1, characterized in that: In step 1, the preprocessing process may adopt commonly used image preprocessing methods including grayscale, denoising, binarization, edge detection, and morphological processing.

3. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 1, characterized in that: In step 2, the instrument image database is a pre-established database that can provide training for the machine learning algorithm, and the database stores pictures and video data of various pointer instruments, float instruments, and liquid level instruments.

4. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 1, characterized in that: In step 2, the machine learning image recognition algorithm uses a combination of an artificial neural network, a support vector machine, and multiple existing algorithms.

5. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 4, characterized in that: In step 3.1, step 3.2, and step 3.3, scale digit recognition is obtained by an algorithm obtained through a convolutional neural network algorithm, and the digit edge range is obtained, and the coordinates of the digit center point are further obtained, providing a basis for subsequent readings.

6. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 5, characterized in that: In step 3.1, step 3.2, and step 3.3, the pointer position recognition, float position recognition, and liquid level position recognition are obtained through Roberts operator, Prewitt operator, Sobel operator, Canny operator, and Laplacian operator edge recognition algorithm, and finally the coordinate data of the pointer tip, the upper edge of the float, and the position of the liquid level are obtained, providing a basis for subsequent readings.

7. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 6, characterized in that: The specific steps for reading the meter using the proportional method in step 3.1 are as follows: (1) based on the coordinates of the center point of each number and the coordinates of the pointer tip; (2) calculating the distance ratio x:y between the needle tip and the two nearest numbers (a and b); (3) obtaining the final reading by calculating the formula (1-x)a / y+xb / y.

8. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 7, characterized in that: In step 3.2 and step 3.3, the specific steps of reading the instrument reading by the intersection method are: (1) the coordinates of the center point of each number and the coordinates of the position of the upper edge of the float or the coordinates of the boundary between the liquid surface and the air; (2) calculating the coordinates of the point where the line connecting the centers of two scale numbers intersects with the upper edge of the float (or the boundary between the liquid surface and the air) or its extension line; (3) calculating the distance ratio x:y between the intersection point and the nearest two numbers (a and b); (3) obtaining the final reading by calculating the formula (1-x)a / y+xb / y.

9. A method for identifying and reading multiple types of mechanical water treatment metering instruments according to claim 1, characterized in that: The specific steps for reading the instrument reading using the scale digital reasoning method in step 3.3 are as follows: (1) based on the coordinates of the center points of each number and the coordinates of the boundary line between the liquid surface and the air; (2) obtaining the coordinates of the intersection point of the line connecting the digital center points and the line where the liquid surface is located; (3) obtaining the two numbers closest to the intersection point (a and b from near to far); (4) obtaining the distance between the two digital center points and the distance from the intersection point to the nearest digital center point (x and y respectively); (5) obtaining the final reading using the calculation formula a-(ba)y / x.

Images