Multi-type industrial instrument identification system suitable for underground engineering and working method thereof

Abstract

The invention discloses a multi-type industrial instrument identification system suitable for underground engineering and a working method thereof, and belongs to the technical field of instrument identification. Comprising an image acquisition module for acquiring and processing multiple types of industrial instrument images and inputting the acquired multiple types of industrial instrument images into an industrial instrument type identification model; the industrial instrument type identification module is used for classifying the multiple types of industrial instrument images acquired by the image acquisition module; the indicator lamp instrument identification module is used for identifying signal lamp signals; the digital instrument identification module is used for obtaining digital instrument data; the pointer instrument identification module is used for identifying the reading of the pointer instrument. The method comprises the following steps: image acquisition; identifying the type of the industrial instrument; and identifying an indicating lamp instrument, a digital instrument and a pointer instrument. According to the method, the image quality is optimized, a solid foundation is laid for subsequent instrument identification work, and the working efficiency and accuracy of identifying multiple types of industrial instruments are improved.

Description

Technical Field

[0001] The invention relates to a multi-type industrial instrument identification system suitable for underground engineering and a working method thereof, belonging to the technical field of instrument identification. Background Art

[0002] Industrial instruments are widely used in various industrial fields due to their low price, huge quantity and wide variety. Accurately reading the values ​​of various industrial instruments is crucial to ensure the smooth progress of industrial production. However, in actual operation, many companies and factories are faced with the challenge of a large number of instruments and a huge workload of reading and identifying.

[0003] Traditionally, the reading and identification of these instruments are mainly done manually. However, this method is not only labor-intensive and costly, but also has a relatively high error rate due to the presence of human factors. In order to reduce the labor intensity of workers and improve the degree of automation of instrument reading and identification, the existing technology introduces automatic instrument identification technology. Although automatic instrument identification technology has alleviated the problem of manual reading to a certain extent, it still faces many challenges in practical applications. Due to the large number of instrument types to be identified and the complex and changeable background environment, existing instrument identification machines are often prone to errors during automatic identification, resulting in low recognition accuracy. This problem not only affects the efficiency and quality of industrial production, but also brings many inconveniences to the normal operation of enterprises.

[0004] Therefore, the present application provides an industrial instrument automatic identification system and method to overcome the deficiencies in the prior art. Summary of the invention

[0005] In order to solve the problems existing in the background technology, the present invention provides a multi-type industrial instrument identification system suitable for underground engineering and a working method thereof.

[0006] To achieve the above object, the present invention adopts the following technical solution: a multi-type industrial instrument identification system suitable for underground engineering, comprising

[0007] An image acquisition module collects and processes images of various types of industrial instruments, and inputs the collected images of various types of industrial instruments into an industrial instrument type recognition model;

[0008] The industrial instrument type recognition module is used to classify the multi-type industrial instrument images collected by the image acquisition module, input the classified indicator instrument images into the indicator instrument recognition module, input the classified digital instrument images into the digital instrument recognition module, and input the classified pointer instrument images into the pointer instrument recognition module;

[0009] Indicator light instrument recognition module, used to identify signal light signals;

[0010] A digital instrument identification module, used to obtain digital instrument data;

[0011] The pointer instrument recognition module is used to identify the indication of the pointer instrument.

[0012] The present invention provides a working method of a multi-type industrial instrument identification system applicable to underground engineering, the method comprising the following steps:

[0013] S1: Image acquisition;

[0014] S101: image acquisition and processing;

[0015] S10101: Use inspection robots to collect images of various types of industrial instruments in industrial plants in sequence;

[0016] S10102: filtering and denoising the collected images of various types of industrial instruments;

[0017] S10103: graying the denoised multi-type industrial instrument images, converting the multi-channel color images into single-channel grayscale images;

[0018] S10104: Continue to perform limited contrast histogram equalization processing on the grayscale image.

[0019] S102: marking of images;

[0020] S10201: Divide all collected images into two groups, the first group is a training set, and the second group is a test set;

[0021] S10202: Mark all images so that each group of images corresponds to three actual industrial instrument types: signal light instrument, digital instrument, and pointer instrument;

[0022] S103: training model;

[0023] S104: Verification and tuning of deep learning models.

[0024] S2: Industrial instrument type identification;

[0025] Use the optimized industrial instrument type recognition model to recognize the acquired multi-type industrial instrument images, and classify the industrial instrument images into indicator instrument images, digital instrument images, and pointer instrument images; input the classified indicator instrument images into the indicator instrument recognition module; input the classified digital instrument images into the digital instrument recognition module; input the classified pointer instrument images into the pointer instrument recognition module;

[0026] S3: indicator light instrument identification;

[0027] S301: Image annotation;

[0028] S302: training an indicator light signal recognition model;

[0029] S303: indicator light signal recognition.

[0030] S4: Digital instrument identification;

[0031] S401: Extraction of instrument digital area

[0032] S40101: Image annotation;

[0033] S40102: Model training;

[0034] S40103: Extraction of digital area;

[0035] S402: Training of digital detection model

[0036] S40201: Image annotation;

[0037] S40202: Model training;

[0038] S403: instrument indication identification;

[0039] S40301: Input the digital region image into the digital detection model to obtain a target detection result;

[0040] S40302: Export target frame related information. Arrange the detection categories from small to large according to the x-coordinate size of the target frame and convert them into corresponding types of output to obtain digital instrument data.

[0041] S5: Pointer instrument identification.

[0042] S501: dial extraction;

[0043] S50101: Image mean shift blur processing;

[0044] S50102: Image grayscale;

[0045] S50103: Detection and extraction of dial.

[0046] S502: Extracting the pointer and the center of the circle;

[0047] S50201: Image processing;

[0048] S50202: Dial scale line outline extraction;

[0049] S50203: Extraction of the center of the dial pointer;

[0050] S50204: Extraction of pointer endpoints and dial center.

[0051] S503: training of the start scale, end scale and maximum range detection model;

[0052] S50301: Image labeling;

[0053] S50302: Model training;

[0054] S504: Identify the number indicated by the pointer.

[0055] S50401: Maximum range detection;

[0056] S50402: Acquisition of the initial scale and end scale dial positions;

[0057] S50403: Identification of dial readings.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The present invention can effectively blur the background noise and interference items through the processing of multi-type industrial instrument image algorithms, thereby making the contrast between the industrial instrument and the surrounding background more significant. This not only optimizes the image quality, but also lays a solid foundation for subsequent instrument recognition work, and improves the work efficiency and accuracy of identifying multi-type industrial instruments. DETAILED DESCRIPTION

[0060] The technical solution of the present invention is clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0061] A multi-type industrial instrument identification system suitable for underground engineering, including

[0062] An image acquisition module collects and processes images of various types of industrial instruments, and inputs the collected images of various types of industrial instruments into an industrial instrument type recognition model;

[0063] The industrial instrument type recognition module is used to classify the multi-type industrial instrument images collected by the image acquisition module, input the classified indicator instrument images into the indicator instrument recognition module, input the classified digital instrument images into the digital instrument recognition module, and input the classified pointer instrument images into the pointer instrument recognition module;

[0064] Indicator light instrument recognition module, used to identify red, yellow and green signal lights;

[0065] A digital instrument identification module, used to obtain digital instrument data;

[0066] The pointer instrument recognition module is used to identify the indication of the pointer instrument.

[0067] The present invention provides a working method of a multi-type industrial instrument identification system applicable to underground engineering, the method comprising the following steps:

[0068] S1: Image acquisition;

[0069] S101: image acquisition and processing;

[0070] S10101: Use inspection robots to collect images of various types of industrial instruments in industrial plants in sequence;

[0071] S10102: filtering and denoising the collected images of various types of industrial instruments;

[0072] Due to the intensity of the light source and the heat of the camera during the image acquisition process, the image will produce noise. Here, Gaussian filtering is used to deal with the noise problem during image acquisition.

[0073] S10103: graying the multi-type industrial instrument images after noise reduction using a weighted average graying method, and converting the multi-channel color images into single-channel gray images;

[0074] S10104: The grayscale image is further subjected to limited contrast histogram equalization processing to enhance the image contrast, thereby further highlighting the industrial instrument while suppressing the enhancement effect of background noise as much as possible.

[0075] S102: marking of images;

[0076] S10201: Divide all collected images into two groups according to the ratio of 7:3, the first group is the training set and the second group is the test set;

[0077] S10202: Mark all images so that each group of images corresponds to three actual industrial instrument types: signal light instrument, digital instrument, and pointer instrument;

[0078] S103: training model;

[0079] S10301: Based on deep learning technology, the training set is used as model training data; a VGG-16 convolutional neural network is constructed, and training sets of different industrial instrument types are input to extract training set features;

[0080] S10302: Use deep learning technology CNN for supervised learning and data training to obtain an industrial instrument type recognition model;

[0081] S104: Verification and tuning of deep learning models;

[0082] S10401: input the test set as an input parameter into the industrial instrument type recognition model;

[0083] S10402: Based on the verification results, the industrial instrument type identification model, training weight parameters and network topology parameters are optimized to obtain the industrial instrument type identification model.

[0084] S2: Industrial instrument type identification;

[0085] Use the optimized industrial instrument type recognition model to recognize the acquired multi-type industrial instrument images, and classify the industrial instrument images into indicator instrument images, digital instrument images, and pointer instrument images; input the classified indicator instrument images into the indicator instrument recognition module; input the classified digital instrument images into the digital instrument recognition module; input the classified pointer instrument images into the pointer instrument recognition module;

[0086] S3: indicator light instrument identification;

[0087] S301: Image labeling: Use LabelImg to label the red, yellow, and green indicator lights corresponding to the indicator light instrument to obtain an indicator light detection image set;

[0088] S302: Training of indicator light signal recognition model: inputting the detection image set into the YOLOv8 model for training to obtain the indicator light signal recognition model;

[0089] S303: indicator light signal recognition: input the indicator light instrument image obtained by classification into the indicator light signal recognition model to recognize the three types of signal lights: red, yellow and green.

[0090] Generally, a red signal light means that the equipment has stopped working, a green signal light means that the equipment is working normally, and a yellow signal light means that the equipment is faulty. When a yellow signal light is identified, a fault report is made.

[0091] S4: Digital instrument identification;

[0092] S401: Extraction of digital area of ​​instrument

[0093] S40101: Image annotation: Use LabelImg to annotate the digital area of ​​the digital instrument image and obtain a detection image set of the digital area of ​​the instrument;

[0094] S40102: Model training: Input the detection image set into the YOLOv8 model for training to obtain the instrument digital area detection model;

[0095] S40103: Extraction of digital area: Input the digital instrument image into the instrument digital area detection model to obtain the digital area target detection frame and extract it.

[0096] S402: Training of digital detection model

[0097] S40201: Image annotation: Use LabelImg to annotate the numbers "0-9" and decimal points as well as common units such as "m, k, A, V, Ω, M, W, h", and obtain the data set;

[0098] S40202: Model training: Input the data set into the YOLOv8 model for training to obtain a digital detection model;

[0099] S403: instrument indication identification;

[0100] S40301: Input the digital region image into the digital detection model to obtain a target detection result;

[0101] S40302: Export target frame related information. Arrange the detection categories according to the x-coordinate size of the target frame from small to large and convert them into corresponding types, such as strings, and output them to obtain digital instrument data.

[0102] S5: Pointer instrument identification.

[0103] S501: dial extraction;

[0104] S50101: Image mean shift blur processing: Use the pyrMeanShiftFiltering function in OpenCV to perform color smoothing filtering on the original image;

[0105] S50102: Image grayscale: The weighted average grayscale processing method is used to convert the multi-channel color image into a single-channel grayscale image. Further, the COLOR_BGR2GRAY function in OpenCV can be used here to convert the image into a grayscale image.

[0106] S50103: Detection and extraction of dial: Based on Hough transform, circle detection is performed using the HoughCircles function in OpenCV to obtain the center and radius of the dial, and the dial is segmented based on the center and radius.

[0107] S502: Extracting the pointer and the center of the circle;

[0108] S50201: Image processing: Perform Gaussian filtering on the dial after filtering out the background, convert it into a grayscale image and perform binarization operation;

[0109] S50202: Dial scale line contour extraction: Use the findContours function in Opencv to locate the contours of the pointer and scale lines, and filter out the pointer contour and scale line contour based on the distance from the contour center point to the dial center point detected using the Hough circle, and then fit the contour into a straight line to complete the detection of the scale lines;

[0110] S50203: Extraction of the center of the dial pointer: After obtaining all the scale lines, extend them in pairs to obtain all the intersections, and calculate the average of the x-coordinates and y-coordinates of all the intersections. The average is the coordinate of the center point of the dial after optimization;

[0111] S50204: Extraction of pointer endpoints and dial center: Same as the initial step of identifying the center point of the dial. After filtering out the pointer outline, obtain the coordinates of the center point of the pointer outline. Use the probabilistic Hough line transformation function to detect the straight lines in the pointer outline image, filter out the longest straight line segment, obtain the endpoints at both ends of the straight line segment, calculate the distance between the two end points of the straight line segment and the center point of the dial, and the larger distance is the pointer endpoint.

[0112] S503: training of the start scale, end scale and maximum range detection model;

[0113] S50301: Image labeling: Use LabelImg to mark the start and end scales and the maximum range of the dial, and obtain the start scale, end scale and maximum range detection data set.

[0114] S50302: Model training: Input the data set into the YOLOv8 model for training to obtain the detection model of the initial scale, end scale and maximum range of the dial.

[0115] S504: Identify the number indicated by the pointer.

[0116] S50401: Maximum range detection: Input the image into the detection model of the initial scale, the end scale and the maximum range to obtain the maximum range target frame of the dial, segment it out, and identify the maximum range of the dial based on the digital detection model obtained in S40202.

[0117] S50402: Acquisition of the initial scale and end scale dial positions: Input the image into the YOLOv8 detection model to obtain the initial scale and end scale dial target frame position information of the dial, and obtain the coordinates of the start scale line and end scale line of the dial according to the target frame position information.

[0118] S50403: Identification of dial readings: Identify the readings of pointer instruments based on the positions of the dial center, pointer endpoints, starting scale, ending scale, and the maximum range of the dial.

[0119] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the present invention can be implemented in other forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered exemplary and non-restrictive in all respects, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalents of the claims be included in the present invention.

[0120] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A multi-type industrial instrument identification system suitable for underground engineering, characterized by: It includes an image acquisition module, which acquires and processes images of various types of industrial instruments, and inputs the acquired images of various types of industrial instruments into an industrial instrument type recognition model; The industrial instrument type recognition module is used to classify the multi-type industrial instrument images collected by the image acquisition module, input the classified indicator instrument images into the indicator instrument recognition module, input the classified digital instrument images into the digital instrument recognition module, and input the classified pointer instrument images into the pointer instrument recognition module; Indicator light instrument recognition module, used to identify signal light signals; A digital instrument identification module, used to obtain digital instrument data; The pointer instrument recognition module is used to identify the indication of the pointer instrument.

2. A working method of the multi-type industrial instrument identification system suitable for underground engineering according to claim 1, characterized in that: The method comprises the following steps: S1: Image acquisition; S2: Industrial instrument type identification; Use the optimized industrial instrument type recognition model to recognize the acquired multi-type industrial instrument images, and classify the industrial instrument images into indicator instrument images, digital instrument images, and pointer instrument images; input the classified indicator instrument images into the indicator instrument recognition module; Inputting the classified digital instrument image into a digital instrument recognition module; The classified pointer instrument image is input into the pointer instrument recognition module; S3: indicator light instrument identification; S4: Digital instrument identification; S5: Pointer instrument identification.

3. The method according to claim 2, characterized in that: The S1 comprises the following steps: S101: image acquisition and processing; S102: marking of images; S10201: Divide all collected images into two groups, the first group is a training set, and the second group is a test set; S10202: Mark all images so that each group of images corresponds to three actual industrial instrument types: signal light instrument, digital instrument, and pointer instrument; S103: training model; S104: Verification and tuning of deep learning models.

4. The method according to claim 3, characterized in that: The S101 includes the following steps: S10101: Use inspection robots to collect images of various types of industrial instruments in industrial plants in sequence; S10102: filtering and denoising the collected images of various types of industrial instruments; S10103: graying the denoised multi-type industrial instrument images, converting the multi-channel color images into single-channel grayscale images; S10104: Continue to perform limited contrast histogram equalization processing on the grayscale image.

5. The method according to claim 2, characterized in that: The S3 comprises the following steps: S301: Image annotation; S302: training an indicator light signal recognition model; S303: indicator light signal recognition.

6. The method according to claim 2, characterized in that: The S4 comprises the following steps: S401: Extraction of instrument digital area S40101: Image annotation; S40102: Model training; S40103: Extraction of digital area; S402: Training of digital detection model S40201: Image annotation; S40202: Model training; S403: instrument indication identification; S40301: Input the digital region image into the digital detection model to obtain a target detection result; S40302: Export target frame related information. Arrange the detection categories from small to large according to the x-coordinate size of the target frame and convert them into corresponding types of output to obtain digital instrument data.

7. The method according to claim 2, characterized in that: The S5 comprises the following steps: S501: dial extraction; S502: Extracting the pointer and the center of the circle; S503: training of the start scale, end scale and maximum range detection model; S50301: Image labeling; S50302: Model training; S504: Identify the number indicated by the pointer.

8. The method according to claim 7, characterized in that: The S501 includes the following steps: S50101: Image mean shift blur processing; S50102: Image grayscale; S50103: Detection and extraction of dial.

9. The method according to claim 8, characterized in that: The S502 includes the following steps: S50201: Image processing; S50202: Dial scale line outline extraction; S50203: Extraction of the center of the dial pointer; S50204: Extraction of pointer endpoints and dial center.

10. The method according to claim 9, characterized in that: The S504 includes the following steps: S50401: Maximum range detection; S50402: Acquisition of the initial scale and end scale dial positions; S50403: Identification of dial readings.