Digital multimeter identification method

Abstract

The invention discloses a digital multimeter identification method, which comprises the following steps of: acquiring a background image of a digital multimeter which is not started; the method comprises the following steps: acquiring an image when a signal is input after startup, displaying numbers on a dial plate at the moment, and performing segmentation and color conversion on the digital image to obtain a grayscale image; determining whether the grayscale image of the digital image contains a background image, and subtracting the acquired image to obtain a background-free character image; carrying out two-dimensional Fourier transform on the digital image to obtain a spectrogram, analyzing a local peak value trend on the spectrogram to obtain an inclination angle of a character, carrying out shear correction on the digital image according to the angle, and changing the image into a character image without inclination; the position of each character is determined through raster scanning, the position of each stroke segment of each character is determined, and each number is recognized through a seven-segment code method. The method for obtaining the character inclination angle through two-dimensional Fourier transform is high in speed, good in stability and easy to program and implement.

Description

Technical Field

[0001] The invention relates to the field of digital multimeter identification, and in particular to a digital multimeter identification method. Background Art

[0002] With the continuous development of the times, electronic and information technology have made rapid and tremendous progress, and many traditional industries have gradually realized digitalization, intelligence and automation. Ordinary digital multimeters are widely used in various industries and fields. In order to ensure the accuracy of gas measurement, regular verification is required. The usual digital multimeter calibration is manual operation. With the advancement of technology, it is developing towards automatic calibration. Ordinary digital multimeters are cheap and have no digital interface. Automatic calibration requires the help of machine vision technology. At this time, it is necessary to collect the image of the numbers displayed on the dial, identify the values ​​with the help of image recognition algorithms, and then obtain the measurement uncertainty by comparing with the standard values ​​to realize the function of automatic calibration.

[0003] The character recognition algorithms and software of the automatic calibration systems of digital multimeters on the market currently still have problems in improving reliability and speed, and cannot fully meet the high standards of users. More optimized algorithms are needed to improve the system and achieve mature automatic calibration. Summary of the invention

[0004] In view of one or more problems in the prior art, a digital multimeter identification method and inspection system are proposed, which can inspect whether the digital multimeter is a background image.

[0005] In one aspect of the present invention, a digital multimeter identification method is provided, comprising:

[0006] S1, using a high-definition camera to obtain a digital image of a digital multimeter, and collecting a background image of the digital multimeter when it is not turned on;

[0007] S2, collecting an image after the device is turned on with a signal input, at which time the dial displays numbers, and performing segmentation and color conversion on the digital image to obtain a grayscale image;

[0008] S3, using a background image detection algorithm to check the grayscale image to determine whether the grayscale image of the digital image contains a background image, subtracting the collected images to obtain a character image without background;

[0009] S4, performing a two-dimensional Fourier transform on the digital image to obtain a spectrum diagram, analyzing the local peak trend on the spectrum diagram to obtain the tilt angle of the character, and performing misalignment correction on the digital image according to the angle to convert the image into a non-tilted character image;

[0010] S5. Determine the position of each character through raster scanning, and determine the position of each stroke segment of each character, and use the seven-segment code method to identify each digit, thereby identifying the entire digit.

[0011] Preferably, the step of checking the grayscale image using a background image detection algorithm includes: extracting all object contour data and texture data from the grayscale image; performing feature comparison on all object contour data and texture data of the grayscale image using a background image detection algorithm created based on the contour features and texture features of all detected objects to determine whether the grayscale image of the digital image contains a background image.

[0012] Preferably, the grayscale image is checked multiple times using a background image detection algorithm.

[0013] Preferably, the step of segmenting and color transforming the digital image comprises: detecting a high-noise area and a low-noise area in the digital image; and excluding the high-noise area and the low-noise area from the grayscale image.

[0014] Preferably, the method further includes the step of: expanding the database by multiple feature extractions using scanned images of known digital multimeter categories, which must include a background image and a scanned image that is similar to the background image but not the background image, to increase the accuracy of the background image detection algorithm.

[0015] Preferably, the process of manual database expansion includes the shape and size of the background image in the database expansion image.

[0016] Preferably, the background image detection algorithm is obtained by extracting the contour features and texture features of all detected objects in the sample using the Canny algorithm, and the extraction process is related to the establishment of the Gaussian function and the calculation of the gradient amplitude and gradient direction.

[0017] Preferably, the method further includes the step of: for digital multimeters that cannot be detected by the background image detection algorithm, and digital multimeters that are detected as background images but are not actually background images, expanding the database through multiple feature extractions to increase the accuracy of background image detection algorithm recognition.

[0018] Preferably, the background image detection algorithm is established for different sizes of background images.

[0019] In another aspect of the present invention, the method is implemented by different modules, including:

[0020] An image acquisition module, which uses a high-definition camera to acquire a digital image of the digital multimeter and acquires a background image when the digital multimeter is not turned on;

[0021] An image preprocessing module, which collects an image after the device is turned on and a signal is input, when the dial displays numbers, and performs segmentation and color conversion on the digital image to obtain a grayscale image;

[0022] A character image acquisition module, which uses a background image detection algorithm to check the grayscale image to determine whether the grayscale image of the digital image contains a background image, and subtracts the collected images to obtain a character image without background;

[0023] The spectrum analysis module performs a two-dimensional Fourier transform on the digital image to obtain a spectrum, analyzes the local peak trend on the spectrum, obtains the tilt angle of the character, and performs mis-cut correction on the digital image according to the angle to convert the image into a non-tilted character image;

[0024] The digital recognition module determines the position of each character through raster scanning, and determines the position of each stroke segment of each character, and uses the seven-segment code method to recognize each digit, thereby recognizing the entire digit.

[0025] Beneficial effects:

[0026] The present invention proposes a digital multimeter recognition method, which can automatically perform background image inspection, improve the inspection efficiency and accuracy. The present invention is a good improvement and optimization of the existing method, and has a good improvement effect on the existing algorithm. Compared with the existing algorithm, it has the advantages of strong anti-noise ability, high recognition accuracy, and fast recognition. The present invention adopts a method of subtracting the background image from the target image to obtain a useful target image, which reduces the range of useful images and speeds up the processing speed of the system. The present invention adopts a method of obtaining the tilt angle of characters by two-dimensional Fourier transform, which has high speed, good stability, and is easy to program and implement. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a schematic flow chart of the method of the present invention;

[0028] Figure 2 It is a schematic diagram of the composition of the modules for implementing the method of the present invention. DETAILED DESCRIPTION

[0029] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not have to be used to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, materials or methods are not specifically described.

[0030] Reference throughout this specification to "one embodiment," "an embodiment," "one example," or "example" means that a particular feature, structure, or characteristic described in connection with the embodiment or example is included in at least one embodiment of the present invention.

[0031] Therefore, the phrases "in one embodiment," "in an embodiment," "one example," or "example" that appear in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, the particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0032] like Figure 1 As shown, a method for identifying a digital multimeter comprises:

[0033] S1, using a high-definition camera to obtain a digital image of a digital multimeter, and collecting a background image of the digital multimeter when it is not turned on;

[0034] S2, collecting an image after the device is turned on with a signal input, at which time the dial displays numbers, and performing segmentation and color conversion on the digital image to obtain a grayscale image;

[0035] S3, using a background image detection algorithm to check the grayscale image to determine whether the grayscale image of the digital image contains a background image, subtracting the collected images to obtain a character image without background;

[0036] S4, performing a two-dimensional Fourier transform on the digital image to obtain a spectrum diagram, analyzing the local peak trend on the spectrum diagram to obtain the tilt angle of the character, and performing misalignment correction on the digital image according to the angle to convert the image into a non-tilted character image;

[0037] S5. Determine the position of each character through raster scanning, and determine the position of each stroke segment of each character, use the seven-segment code method to identify each digit, and thus identify the entire digit.

[0038] Use a high-definition camera to shoot the digital multimeter and collect image information on the multimeter screen. This step is divided into two parts:

[0039] When the DMM is not turned on, a background image is acquired as a comparison benchmark. This background image can be used to remove background interference later.

[0040] The background image will be used to separate the image to facilitate further extraction of effective information.

[0041] After the digital multimeter is turned on, the dial will display certain digital information, representing the measurement results.

[0042] After acquiring the image after power-on, perform the following processing on the image:

[0043] Image segmentation: Extract the digital area displayed by the multimeter from the original image and remove other redundant areas (such as interference information such as the dial shell).

[0044] Color transformation: Convert the extracted digital region image into a grayscale image (i.e. black and white image) to facilitate subsequent processing. Grayscale images simplify the color dimension and enhance computational efficiency.

[0045] Check whether the current grayscale image still contains elements of the background image (for example, problems such as light reflection and screen reflection may cause background residue).

[0046] The background image detection algorithm specifically performs the following operations:

[0047] Perform pixel-by-pixel comparison or subtraction between the grayscale image and the previously acquired background image to eliminate background interference.

[0048] The image retained after processing is a character image without background, which only contains digital information.

[0049] Extract the spectrum information of digital images using two-dimensional Fourier transform:

[0050] Fourier transform can convert images from spatial domain to frequency domain, and the spectrum diagram reflects the directional information and periodic characteristics in the image.

[0051] Analyze the local peak trend in the spectrum graph to determine whether the character image has a tilt problem (such as manual shooting or display screen angle problem).

[0052] After calculating the tilt angle of the image, correct the image:

[0053] Shear correction is a geometric transformation method used to adjust a tilted image to a horizontal or vertical state to make the characters legible.

[0054] Use the raster scanning method to scan the image row by row or column by column to locate the area where each character is located (character bounding box).

[0055] Raster scanning is to read the image line by line at certain intervals, analyze the changes in pixel values, and identify the character area.

[0056] Perform position detection on each stroke of each character:

[0057] The geometric features of the characters are extracted by performing nonlinear fitting (such as spline fitting or Bezier curve fitting) on ​​the strokes of the characters.

[0058] After feature extraction, the strokes are matched with predefined character models to gradually recognize each digit.

[0059] Combine all the recognized numbers to form a complete recognition result.

[0060] In order to obtain a useful signal with a high signal-to-noise ratio, the present invention adopts a method of first collecting a background header image before the machine is turned on, then collecting a header image containing characters after the machine is turned on, and then subtracting the two to obtain a character image with a high signal-to-noise ratio.

[0061] Collect the image after the power-on with signal input, when the dial displays numbers. Subtract the collected images to obtain the character image without background.

[0062] Perform a two-dimensional Fourier transform on the digital image to obtain a spectrum. Analyze the local peak trend on the spectrum to obtain the tilt angle of the character. Perform mis-cut correction on the digital image according to the angle to convert the image into a non-tilted character image.

[0063] The character display of the digital multimeter is realized by seven-segment code, and the character recognition is also based on the seven-segment code. The character is decomposed into seven separate segments, each segment is numbered, and each number corresponds to a string of seven-bit codes. For example, the seven-bit code corresponding to the number "1" is "0010010", the seven-bit code corresponding to the number "2" is "0111101", and the seven-bit code corresponding to the number "8" is "1111111".

[0064] In the previous step, the position of each character and character segment has been obtained. According to the pixel value in each character segment, its code can be determined, and the value of the character can be determined according to the code, thereby identifying the values ​​of all characters.

[0065] The step of checking the grayscale image using a background image detection algorithm comprises:

[0066] Extracting all object contour data and texture data from the grayscale image;

[0067] A background image detection algorithm created based on the contour features and texture features of all detected objects is used to perform feature comparison on all object contour data and texture data of the grayscale image to determine whether the grayscale image of the digital image contains a background image.

[0068] Specifically, all object contour data and texture data are extracted from the grayscale image, and then the background image detection algorithm created based on the contour features and texture features of all detected objects is used to perform feature comparison on all object contour data and texture data of the grayscale image to determine whether the grayscale image of the digital image contains a background image. The process of feature extraction, algorithm establishment and algorithm update is described in detail below.

[0069] Image feature extraction is the first step in object feature comparison and detection. Low-level feature extraction generally includes: interest point detection-based and dense extraction-based methods. Based on interest point detection, such as Harris corner detector, FAST (Features from Accelerated Segment Test) operator, LoG (Laplacian of Gaussian), DoG (Difference of Gaussian), etc. Based on dense extraction (dense), all features are generally used, such as SIFT (Scale-invariant feature transform), HOG (Histogram of Oriented Gradient), LBP (Local Binary Pattern), etc. Due to the great diversity of content in digital images, using a single feature representation cannot meet actual needs. For example, the HOG feature describes the gradient direction density distribution of all image gradients or edges, which can be considered as a description of the appearance and contour of all detected objects. LBP can be considered to describe a kind of texture data. In the application of object detection, contour information is more useful than texture data, but the two can complement each other. In a digital image, when DMMs are densely stacked together, the contour information is no longer obvious in the image, but the texture features are obvious; when a large DMM is placed alone, the texture data is not obvious, but the contour information is obvious. In order to cope with the diversity of DMM types, in an embodiment of the present invention, the underlying features use a HOG-LBP adaptive fusion feature (Boosted HOG-LBP) to describe the digital image.

[0070] The area with dense digital multimeters has a small regional conditional entropy value, so when the digital multimeters are dense, the proportion of texture features is larger.

[0071] In addition, in order to improve the detection performance, in the embodiment of the present invention, mid-level features, namely image slices, can be used to describe the most original information of the image. Such mid-level features can be obtained by clustering a large number of image slices.

[0072] The training of feature matching algorithm (two-feature matching is detection, also referred to as detection algorithm below) requires a large number of positive samples that have been expanded by the database (negative samples do not need database expansion). At present, the database expansion of positive samples mostly adopts the method of manual database expansion. Although there are some algorithms and software to assist in database expansion, it is impossible to completely free people. The method of weak database expansion can reduce the degree of human participation, making the database expansion process simple and efficient. In the embodiment of the present invention, two methods are used to expand the database: 1) Use a rectangular box to mark the approximate shape of the background image in the image, and use a variable to describe the size of the background image. The shape and posture can be used as hidden variables when training the algorithm; 2) "Positioning + detection" two steps to achieve weak database expansion.

[0073] The "positioning + detection" method belongs to a weak database expansion algorithm, that is, given a digital image sample, it only needs to tell whether there is a background image in this image and how the background image bottles are placed. In the actual transportation of digital multimeters, the background image bottles are generally stacked upright in the container. There are also cases of stacking horizontally and upside down, but the probability of occurrence is low. According to the actual situation, the sizes of background image bottles are divided into five types: A upright, B inverted, C horizontally perpendicular to the ray gradient direction, D horizontally parallel to the ray gradient direction 1, and E horizontally parallel to the ray gradient direction 2. In a digital image with a known background image bottle size, the background image bottles are regularly stacked together. Irregular placement is very rare in the transportation of large quantities of background image bottles. At this time, the image can be densely divided into image patches of equal size, and then clustered using a clustering algorithm. Since K-means uses distance measurement to cluster visually dissimilar image pieces into one category, in order to solve this problem, the patent uses a detection clustering method to deal with it: first use K-means clustering, then use linear SVM to train a feature comparator for each cluster, and detect them in the image respectively. The class with the most detection results is the positive sample, and the detected area is the area where the background image is located. After the "positioning + monitoring" process, the area of ​​the background image in the digital image can be obtained, and the size of the background image can be added to process it in the same way as the first database expansion method. In the specific implementation of the present invention, any one of the two methods can be selected to deal with the positive sample database expansion problem.

[0074] In the previous article, the approximate area and posture category of the background image were obtained through weak database expansion. With this information, the detection algorithm can be trained. For background images with bottle packaging similar to the background image, five component discriminants can be trained to detect background images of five sizes. Combining the five components together to form the overall feature comparison detection algorithm is the Mult-Component detection algorithm.

[0075] The grayscale image is checked several times using a background image detection algorithm.

[0076] The steps of segmenting and color conversion of digital images include:

[0077] Detecting a multi-noise area and a noise-free area in the digital image;

[0078] Noisy areas and noise-free areas are excluded from the grayscale image.

[0079] Through the scanned images of known digital multimeter categories, which must include background images and scanned images that are similar to the background images but not the background images, the database is expanded through multiple feature extractions to increase the accuracy of background image detection algorithm recognition;

[0080] The process of database expansion includes the form and size of the background image in the database expansion image.

[0081] The background image detection algorithm is to extract the contour features and texture features of all detected objects in the sample using the Canny algorithm. The extraction process is related to the establishment of the Gaussian function and the calculation of the gradient amplitude and gradient direction.

[0082] The Canny algorithm uses a two-dimensional zero-mean Gaussian function and performs a convolution operation on the image matrix to eliminate noise and smooth the image. The expression of the selected Gaussian function is:

[0083]

[0084] The smoothed image expression I(x,y) is:

[0085] I(x,y)=G(x,y)*f(x,y)

[0086] Among them, f(x,y) represents the function of the original image, σ represents the Gaussian filter distribution parameter, x and y represent the horizontal and vertical coordinates of the pixel points in the image, exp represents the exponential function with a natural constant as the base, and π represents pi;

[0087] Gradient amplitude, expressed as:

[0088]

[0089] Gradient direction, expressed as:

[0090]

[0091] Among them, E x (i, j) represents the first-order partial derivative in the x direction, E y(i, j) represents the first-order partial derivative in the y direction, A(i, j) represents the gradient amplitude, Indicates the gradient direction.

[0092] The background image detection algorithm is established for different sizes of background images.

[0093] For background images that cannot be detected by the background image detection algorithm, and digital multimeters that are detected as background images by the background image detection algorithm but are not actually background images, the database is expanded through multiple feature extractions to increase the accuracy of the background image detection algorithm's recognition.

[0094] A digital multimeter identification method is implemented by the following modules:

[0095] An image acquisition module, which uses a high-definition camera to acquire a digital image of the digital multimeter and acquires a background image when the digital multimeter is not turned on;

[0096] An image preprocessing module, which collects an image after the device is turned on and a signal is input, when the dial displays numbers, and performs segmentation and color conversion on the digital image to obtain a grayscale image;

[0097] A character image acquisition module, which uses a background image detection algorithm to check the grayscale image to determine whether the grayscale image of the digital image contains a background image, and subtracts the collected images to obtain a character image without background;

[0098] The spectrum analysis module performs a two-dimensional Fourier transform on the digital image to obtain a spectrum, analyzes the local peak trend on the spectrum, obtains the tilt angle of the character, and performs mis-cut correction on the digital image according to the angle to convert the image into a non-tilted character image;

[0099] The digital recognition module determines the position of each character through raster scanning, and determines the position of each stroke segment of each character, and uses the seven-segment code method to recognize each digit, thereby recognizing the entire digit.

[0100] The present invention extracts all object contour data and texture data from the grayscale image, and uses a background image detection algorithm created based on the contour features and texture features of all detected objects to perform feature comparison on all object contour data and texture data of the grayscale image to determine whether the grayscale image of the digital image contains a background image.

[0101] The solution of the above embodiment can perform background image detection on the digital multimeter, especially the container scanning image, and can provide intelligent assistance to the staff.

[0102] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for identifying a digital multimeter, characterized in that: The method includes: S1, using a high-definition camera to obtain a digital image of a digital multimeter, and collecting a background image of the digital multimeter when it is not turned on; S2, collecting an image after the device is turned on with a signal input, at which time the dial displays numbers, and performing segmentation and color conversion on the digital image to obtain a grayscale image; S3, using a background image detection algorithm to check the grayscale image to determine whether the grayscale image of the digital image contains a background image, subtracting the collected images to obtain a character image without background; S4, performing a two-dimensional Fourier transform on the digital image to obtain a spectrum diagram, analyzing the local peak trend on the spectrum diagram to obtain the tilt angle of the character, and performing misalignment correction on the digital image according to the angle to convert the image into a non-tilted character image; S5. Determine the position of each character through raster scanning, and determine the position of each stroke segment of each character, use the seven-segment code method to identify each digit, and thus identify the entire digit.

2. A digital multimeter identification method as claimed in claim 1, characterized in that: The step of using a background image detection algorithm to check the grayscale image includes: Extracting all object contour data and texture data from the grayscale image; A background image detection algorithm created based on the contour features and texture features of all detected objects is used to perform feature comparison on all object contour data and texture data of the grayscale image to determine whether the grayscale image of the digital image contains a background image.

3. A digital multimeter identification method as claimed in claim 1, characterized in that: The grayscale image is checked multiple times using a background image detection algorithm.

4. A digital multimeter identification method as claimed in claim 1, characterized in that: The steps of segmenting and color transforming the digital image include: Detecting a multi-noise area and a noise-free area in the digital image; Noisy areas and noise-free areas are excluded from the grayscale image.

5. A digital multimeter identification method as claimed in claim 1, characterized in that: Also includes the steps: Through the scanned images of known digital multimeter categories, which must include background images and scanned images that are similar to the background images but not the background images, the database is expanded through multiple feature extractions to increase the accuracy of background image detection algorithm recognition; The process of database expansion includes the form and size of the background image in the database expansion image.

6. A method for identifying a digital multimeter as claimed in claim 5, characterized in that: The background image detection algorithm is obtained by extracting the contour features and texture features of all detected objects in the sample using the Canny algorithm. The extraction process is related to the establishment of the Gaussian function and the calculation of the gradient amplitude and gradient direction.

7. A method for identifying a digital multimeter as claimed in claim 5, characterized in that: The Canny algorithm uses a two-dimensional zero-mean Gaussian function and performs a convolution operation on the image matrix to eliminate noise and smooth the image. The expression of the selected Gaussian function is: The smoothed image expression I(x,y) is: I(x,y)=G(x,y)*f(x,y) Among them, f(x,y) represents the function of the original image, σ represents the Gaussian filter distribution parameter, x and y They are respectively represented as the horizontal and vertical coordinates of the pixel points in the image, exp represents the exponential function with a natural constant as the base, and π represents pi; Gradient amplitude, expressed as: Gradient direction, expressed as: Among them, E x (i, j) represents the first-order partial derivative in the x direction, E y (i, j) represents the first-order partial derivative in the y direction, A(i, j) represents the gradient amplitude, Indicates the gradient direction.

8. A method for identifying a digital multimeter as claimed in claim 5, characterized in that: The background image detection algorithm is established for background images of different sizes.

9. A digital multimeter identification method as claimed in claim 1, characterized in that: Also includes the steps: For background images that cannot be detected by the background image detection algorithm, and digital multimeters that are detected as background images by the background image detection algorithm but are not actually background images, the database is expanded through multiple feature extractions to increase the accuracy of the background image detection algorithm's recognition.

Images