Dial plate identification method and device based on image processing
Through the image processing method, the traditional pointer-type dial is automatically recognized, which solves the problems of unreachability, poor reliability and inefficiency of traditional reading methods, and achieves efficient and accurate data reading.
Patent Information
- Application Number
- CN202411814261.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-10
- Publication Date
- 2025-04-18
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional pointer-type dial data reading has problems such as unreachability, poor reliability and inefficiency, especially in extreme environments and in confidential situations.
The dial recognition method based on image processing is used to divide the dial into four categories: semicircular watch + digital watch rectangular watch, two-scale thick pointer circular watch, whole watch and multi-pointer watch, respectively, and different image recognition algorithms are used to identify pointers or numbers.
Improves the efficiency and accuracy of dial recognition, reduces human error, is suitable for various environments, and supports automated production.
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing, and in particular to a dial recognition method and device based on image processing. Background Art
[0002] With the rapid progress of my country's industrial development, more and more factories are gradually turning to automation. However, traditional old-fashioned factories have many problems in the process of transformation. The mechanical instruments used in the factory are all traditional pointer instruments. The reading of data on the instrument often relies on manual reading by on-site personnel. This reading method has great problems:
[0003] (1) Inaccessibility. Traditional instruments are usually installed in places that are invisible to the human eye, making it inconvenient to read data. In addition, in some extreme environments, such as those with corrosive gases, high temperatures, and flammable and explosive materials, it is difficult for workers to get close, which greatly increases the difficulty of data reading.
[0004] (2) Poor reliability. Manual data reading is a highly repetitive and subjective job. There will be certain deviations in the data read by workers at different times and in different states. If the dial data is confidential, there may be a risk of confidentiality.
[0005] (3) Inefficiency. Industrial production operations require efficient completion. Manual reading is undoubtedly a very inefficient behavior and is not conducive to the factory's fully automated production.
[0006] In recent years, with the rapid development of computer vision technology, image processing technology has gradually been applied in all walks of life. In view of the current situation of manual reading of data on traditional pointer dials, it is of great significance to apply image processing technology to the reading of data on traditional pointer dials. The development of a universal automatic reading system for pointer instruments and digital instruments is an urgent need for the development of contemporary industry. Summary of the invention
[0007] The present application provides a dial recognition method and device based on image processing, which are used to solve the problems of low efficiency and low accuracy of existing dial recognition.
[0008] In view of this, the present application provides a dial recognition method based on image processing, comprising:
[0009] Based on the characteristics of the dial, it is divided into four categories through feature integration: (1) semicircular dial + digital dial rectangular dial; (2) round dial with thick pointers on two scales; (3) full dial; (4) multi-pointer dial. For the four different types of dials, different image recognition algorithms are used to identify the pointers or numbers on the dial.
[0010] For a rectangular dial with a semicircular dial + digital dial type, that is, the dial has both a semicircular pointer area and a digital dial area, the identification method includes:
[0011] For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the rectangular table detection is performed to obtain the digital table. For the digital table, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. For the pointer area, the area of the pointer table is first determined by matrix mapping, and then the Hough transform method of finding a straight line is used to determine the position of the pointer. For the found pointer, the fine pointer reading method is used, that is, the distance method is used to identify the fine pointer reading, and the instrument reading is calculated by the distance relationship between the pointer straight line and the adjacent scale line to obtain the pointer recognition result.
[0012] For a round dial with thick pointers and only two scales, the identification method includes:
[0013] For the acquired dial image, we first perform image graying, image defogging, image filtering, and image enhancement preprocessing algorithms in sequence. Then, we perform circular table detection based on Hough transform. In the obtained circular table, we obtain all the contours of the dial by marking the connected areas, marking the strokes, and marking the contours. The largest contour is the location of the pointer. Finally, we determine whether the current state of the pointer is open or closed by the angle of the pointer.
[0014] For a round dial whose dial type is a full dial, the identification method includes:
[0015] For the acquired dial image, we first perform image graying, image dehazing, image filtering, and image enhancement preprocessing algorithms in sequence. Then, we perform circular dial detection based on Hough transform to detect the circular dial in the image. Then, we use regional contrast, regional spatial relationship, center prior saliency features, and correlation matrix algorithm to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to perform the reading recognition of the pointer on the dial.
[0016] For a circular dial with multiple pointers, the identification method includes:
[0017] For the acquired dial image, firstly, the image grayscale, image defogging, image filtering, and image enhancement preprocessing algorithms are executed in sequence, and then the circular dial is detected to detect the circular dial in the image, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image contains a square dial, firstly, the pointer area is determined by the matrix mapping method, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition.
[0018] The present application also provides a dial recognition device based on image processing, comprising:
[0019] Image processing unit. Based on the characteristics of the dial, the image processing unit divides it into four categories: (1) semicircular dial + digital dial rectangular dial; (2) round dial with two thick pointers; (3) full dial; (4) multi-pointer dial. For the four different types of dials, different image recognition algorithms are used to identify the pointers or numbers on the dial.
[0020] Rectangular dial recognition unit for semicircular meter + digital meter. For the acquired dial image, firstly, perform image graying, image defogging, image filtering, and image enhancement preprocessing algorithms in sequence, and then perform rectangular meter detection to obtain a digital meter. For the digital meter, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. For the pointer area, first use matrix mapping to determine the area of the pointer meter, and then use the Hough transform to find the straight line method to determine the pointer position. For the pointer found, the fine pointer reading method is used, that is, the distance method is used to identify the fine pointer reading, and the instrument reading is calculated through the distance relationship from the pointer straight line to the adjacent scale line to obtain the pointer recognition result.
[0021] The thick pointer circular dial recognition unit with only two scales. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the circular table detection based on Hough transform is performed. In the obtained circular table, all the contours of the dial are obtained by marking the connected area, marking the stroke, and marking the contour. The largest contour is found to be the location of the pointer. Finally, the current state of the pointer is determined by the angle of the pointer.
[0022] Whole dial recognition unit. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are executed in sequence, and then the circular dial detection based on Hough transform is executed to detect the circular dial in the image, and then the regional contrast, regional spatial relationship, center prior saliency feature, and correlation matrix algorithm are used to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to perform the reading recognition of the pointer in the dial.
[0023] Multi-pointer recognition unit. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the circular dial is detected. The circular dial in the image is detected, and the pointer sub-image is obtained by using the region growing, ROI region extraction, and refinement algorithms in sequence, and finally the pointer recognition result is obtained; if the image contains a square dial, the pointer area is first determined by the matrix mapping method, and then the pointer sub-image is obtained by using the region growing, ROI region extraction, and refinement algorithms in sequence, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0025] Figure 1 A dial recognition method based on image processing provided in the present application is a schematic diagram of a rectangular dial with a semicircular dial + a digital dial.
[0026] Figure 2 A flow chart of a dial recognition method based on image processing provided in the present application, wherein the dial type is a rectangular dial with a semicircular dial + a digital dial.
[0027] Figure 3 A dial recognition method based on image processing provided in the present application is a schematic diagram of a circular dial with two thick pointers and scales.
[0028] Figure 4 A flow chart of a dial recognition method based on image processing provided in the present application, wherein the dial type is a circular dial with thick pointers and two scales.
[0029] Figure 5 A dial type of a dial recognition method based on image processing provided in this application is a schematic diagram of a whole dial.
[0030] Figure 6 A flow chart of a dial recognition method based on image processing provided in the present application, wherein the dial type is a whole dial.
[0031] Figure 7 A schematic diagram of a dial type of a multi-pointer dial in a dial recognition method based on image processing provided in the present application.
[0032] Figure 8 A flow chart of a dial recognition method based on image processing provided in the present application, wherein the dial type is a multi-pointer dial. DETAILED DESCRIPTION
[0033] The present application proposes a dial recognition method and device based on image processing, which are used to solve the problems of low efficiency and low accuracy of existing dial recognition.
[0034] In order to make the purpose, features, and advantages of the invention of this application more obvious and easy to understand, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described below are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0035] Based on the characteristics of the dial, it is divided into four categories through feature integration: (1) semicircular dial + digital dial rectangular dial; (2) two-scale thick pointer circular dial; (3) full dial; (4) multi-pointer dial. For the four different types of dials, the following image recognition algorithm is used to identify the pointers or numbers on the dial.
[0036] (1) Image preprocessing. For all meters, due to the influence of external factors such as weather and light, the image resolution is not enough, and the subsequent pointer recognition will have large errors. Therefore, image preprocessing is a very critical step.
[0037] Aiming at the problems of noise, illumination and shadow in the identification of pointer instruments, a set of image preprocessing algorithms suitable for automatic identification systems was studied to improve the anti-interference ability of the identification system. In order to ensure the quality of the picture, this application has carried out a series of image preprocessing techniques, such as graying, defogging, filtering operations, image enhancement, binarization, etc. After experimental verification, the final order of image preprocessing is determined as: image graying, image defogging, image filtering, and image enhancement. The image preprocessing operations in this order can ensure the clarity of the image. The use of other orders and other preprocessing algorithms cannot achieve the effect of the current algorithm. After preprocessing, the influence of external environmental factors can be greatly reduced, which is more conducive to the identification of the pointer.
[0038] (2) Rectangular watch detection. The actual acquired image often contains a complex background, so it is necessary to extract the exact watch dial area. The watch dial extraction algorithm is as follows:
[0039] ① Use the Canny algorithm to detect the edges of the instrument image and extract important edge information.
[0040] ② Use Hough transform to detect the straight line information in the edge information and accurately identify the dial area.
[0041] ③ Remove invalid line segments. Based on the prior knowledge of the dial position information, set the limiting conditions to select the useful line segments that can represent the border and filter out other useless line segments. The limiting conditions are as follows: the edge of the square dial should be four approximately horizontal or approximately vertical line segments, and a horizontal line segment must have a common intersection with the other two vertical line segments. Similarly, a vertical line segment must also have a common intersection with the other two horizontal line segments. Invalid line segments are removed through the above limiting conditions.
[0042] ④ Determine the center point of the dial. The center point extracted in this application is not obtained by using the x-axis coordinates and y-axis coordinates of the midpoints of all line segments to obtain the coordinates of the center point of the dial, but by analyzing and processing the approximate horizontal line segments and the approximate vertical line segments separately. That is, the x-axis coordinate of the center point of the dial is obtained by summing and averaging the x-axis coordinates of all approximate horizontal line segments, while the y-axis coordinate is obtained by summing and averaging the y-axis coordinates of all approximate vertical line segments.
[0043] ⑤ Calculate the rectangle equation. For the useful line segments obtained, according to the characteristics of the rectangle, these line segments are divided into four categories, and each category of useful line segments is processed separately and integrated into one line segment, then there will be four line segments, and the intersection of two line segments is used as the point of the rectangle, and four intersection points are obtained. Finally, the rectangle formed by the four intersection points is the final rectangle.
[0044] (3) Digital watch recognition. After the digital watch face is detected by rectangle detection, it is necessary to detect the value of the digital watch face. The digital watch recognition scheme of this application uses the SVM+HOG feature extraction method. First, the features of each number from 0 to 9 are extracted using the HOG feature. The advantage of this feature extraction is that it improves accuracy by calculating densely and evenly spaced grid cells and applying overlapping local contrast normalization, thereby ensuring that the features of each number can be extracted. Then use SVM to classify them. The SVM training process is simple and the classification effect is high. This not only reduces the training time, but also ensures the accuracy of classification.
[0045] (4) Circular dial detection. It can be found through observation that there are many circular shapes on the dial. Therefore, this application needs to detect the circular dial and remove the interference of other external impurities. This application uses Hough transform to find the circle, convert each pixel point to polar coordinates and accumulate the intensity of the center point coordinate of the corresponding circle. In polar coordinates, the intensity value is normalized to between 0-255, and the point with the largest intensity is found, which is the center point of the circle, and the circle is drawn based on this center point.
[0046] (5) Rectangular mapping. Because it is difficult to obtain the complete outline of the large area where the pointer table is located, this application chooses to use the rectangular mapping method. According to the small pointer box of the digital table found, the mapping relationship between the large and small rectangles is found to ensure that the mapping range includes the pointer table, thereby determining the location of the large rectangular box. Because the pointer table is located in a fan-shaped or irregular area, when using rectangular mapping, it is necessary to ensure that the pointer area is included and the mapped area is determined.
[0047] Rectangular mapping is a new concept proposed in this application. This technology is aimed at some meters that contain both digital meters and pointer meters, and sometimes the detected pointer meters are incomplete, or the pointer meters themselves are incomplete. To address this problem, this application proposes a rectangular mapping solution. First, this application finds the position of the digital meter through rectangular detection and determines the coordinates of the four right angles of the digital meter. In the known reference map, the distance between the digital meter and the pointer meter can be known. The coordinates of the four right angles of the pointer meter can be obtained based on these two parameters: the distance between the digital meter and the pointer meter and the coordinates of the four right angles of the digital meter. In this way, it is possible to ensure that a complete pointer meter is obtained, thereby ensuring that the obtained scale is clear and complete, and ensuring the accuracy of meter recognition.
[0048] (6) Finding a connected domain. By observing the thick circular dial with only two scales, it can be found that the area where the pointer is located is the largest connected domain. First, this application needs to use the method of the largest connected domain to determine the exact position of the pointer, because the connected domain formed by the location of the pointer has the largest area. Therefore, this application obtains the location of the pointer by finding the method of the largest connected domain.
[0049] (7) Find the pointer using Hough transform. After finding the pointer dial through rectangular mapping, you can find the relevant information about the pointer, including the straight line equation of the pointer, the position relationship between the pointer and the dot, etc. Identify each scale line within the mapping range, and then find the pointer. As a commonly used straight line detection algorithm, Hough transform has good robustness and can still perform well for straight lines with breakpoints. It is more suitable for pointer recognition and is also an algorithm frequently used in the current automatic reading system of pointer-type instruments. In this application, Hough transform is used as an algorithm for detecting the pointer straight line. Because the length of the pointer is longer than the length of the scale, this application can add this restriction condition: the line segment with the largest line segment length is selected as the pointer line segment, so the position of the pointer can be found. This method is simple to operate, easy to find, and has good results.
[0050] (8) Fine pointer reading identification. In the fine pointer reading identification process, because the pointer is relatively thin, it can be regarded as a straight line. The Hough transform can be used directly to find the straight line, and then the distance method can be used to identify the fine pointer reading, that is, the instrument reading is calculated based on the distance relationship between the pointer straight line and the adjacent scale line.
[0051] (9) Coarse pointer recognition. Coarse pointer recognition is mainly based on regional contrast, regional spatial relationship, central prior saliency features, and correlation matrix algorithm to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to finally obtain the coarse pointer reading recognition.
[0052] (10) Multi-pointer reading. Some dials contain multiple pointers, which makes it very difficult to identify and read the pointers. In order to solve this problem, this application proposes a multi-pointer reading algorithm. First, circular dial detection is performed to detect the circular dial in the image, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained. If the image contains a square dial, the pointer area is first determined by the matrix mapping method, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition.
[0053] See also Figure 1 , it can be found that they have common similarities. For dials that contain both pointer meters and digital meters, this application designs a universal framework to identify the readings of such meters, that is, using image preprocessing + rectangular meter detection + digital meter recognition + rectangular mapping + Hough transform to find the pointer + fine pointer reading recognition for recognition.
[0054] For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the rectangular table detection is performed to obtain the digital table. For the digital table, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. For the pointer area, the area of the pointer table is first determined by matrix mapping, and then the Hough transform method of finding a straight line is used to determine the position of the pointer. For the found pointer, the fine pointer reading method is used, that is, the distance method is used to identify the fine pointer reading, and the instrument reading is calculated by the distance relationship between the pointer straight line and the adjacent scale line to obtain the pointer recognition result.
[0055] Specific identification schemes such as Figure 2 Note: If the pointer meter does not have a rectangular frame, include the entire scale area when mapping with a digital meter.
[0056] See also Figure 3 , it can be found that they have a common similarity, that is, these two circular meters have only two scales: division and addition. For such meters with the same characteristics, this application designs a universal framework to identify the readings of such meters. That is, image preprocessing + circular meter detection + finding connected domains + coarse pointer recognition are used for identification.
[0057] For the acquired dial image, we first perform image graying, image defogging, image filtering, and image enhancement preprocessing algorithms in sequence. Then, we perform circular table detection based on Hough transform. In the obtained circular table, we obtain all the contours of the dial by marking the connected areas, marking the strokes, and marking the contours. The largest contour is the location of the pointer. Finally, we determine whether the current state of the pointer is open or closed by the angle of the pointer.
[0058] Specific identification schemes such as Figure 4 shown.
[0059] See also Figure 5 , it can be found that they have a common similarity, that is, the pointers of these two circular meters are relatively large. For such meters with the same characteristics, this application designs a universal framework to identify the readings of such meters. That is, image preprocessing + circular meter detection + coarse pointer reading recognition is used for identification.
[0060] For the acquired dial image, we first perform image graying, image dehazing, image filtering, and image enhancement preprocessing algorithms in sequence. Then, we perform circular dial detection based on Hough transform to detect the circular dial in the image. Then, we use regional contrast, regional spatial relationship, center prior saliency features, and correlation matrix algorithm to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to perform the reading recognition of the pointer on the dial.
[0061] Specific identification schemes such as Figure 6 shown.
[0062] See also Figure 7 , it can be found that it contains more than one pointer. For such meters with the same characteristics, this application designs a universal framework to identify the readings of such meters. That is, image preprocessing + digital meter recognition + circular meter detection + rectangular mapping + multi-pointer reading recognition is used for recognition.
[0063] For the acquired dial image, firstly, the image grayscale, image defogging, image filtering, and image enhancement preprocessing algorithms are executed in sequence, and then the circular dial is detected to detect the circular dial in the image, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image contains a square dial, firstly, the pointer area is determined by the matrix mapping method, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition.
[0064] Specific identification schemes such as Figure 8 shown.
[0065] The present application also provides a dial recognition device based on image processing technology, comprising:
[0066] Image processing unit. Based on the characteristics of the dial, the image processing unit divides it into four categories: (1) semicircular dial + digital dial rectangular dial; (2) round dial with two thick pointers; (3) full dial; (4) multi-pointer dial. For the four different types of dials, different image recognition algorithms are used to identify the pointers or numbers on the dial.
[0067] Rectangular dial recognition unit for semicircular meter + digital meter. For the acquired dial image, firstly, perform image graying, image defogging, image filtering, and image enhancement preprocessing algorithms in sequence, and then perform rectangular meter detection to obtain a digital meter. For the digital meter, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. For the pointer area, first use matrix mapping to determine the area of the pointer meter, and then use the Hough transform to find the straight line method to determine the pointer position. For the pointer found, the fine pointer reading method is used, that is, the distance method is used to identify the fine pointer reading, and the instrument reading is calculated through the distance relationship from the pointer straight line to the adjacent scale line to obtain the pointer recognition result.
[0068] The thick pointer circular dial recognition unit has only two scales. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the circular table detection is performed. In the obtained circular table, all the contours of the dial are obtained by marking the connected area, marking the stroke, and marking the contour. The largest contour is the location of the pointer. Finally, the angle of the thick pointer is used to determine whether the current pointer state is open or closed.
[0069] Whole dial recognition unit. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are executed in sequence, and then the circular dial detection based on Hough transform is executed to detect the circular dial in the image, and then the regional contrast, regional spatial relationship, center prior saliency feature, and correlation matrix algorithm are used to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to perform the reading recognition of the pointer in the dial.
[0070] Multi-pointer recognition unit. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the circular dial is detected. The circular dial in the image is detected, and the pointer sub-image is obtained by using the region growing, ROI region extraction, and refinement algorithms in sequence, and finally the pointer recognition result is obtained; if the image contains a square dial, the pointer area is first determined by the matrix mapping method, and then the pointer sub-image is obtained by using the region growing, ROI region extraction, and refinement algorithms in sequence, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition.
[0071] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0072] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.
[0073] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0074] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0075] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0076] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A dial recognition method based on image processing, characterized in that: include: According to the characteristics of the dial, they are divided into four categories through feature integration: (1) semicircular dial + digital dial rectangular dial; (2) A round dial with two thick pointers; (3) A full dial; (4) A multi-pointer dial. For the above four different types of dials, different image recognition algorithms are used to identify the pointers or numbers on the dial.
2. The dial recognition method based on image processing according to claim 1, characterized in that: For a rectangular dial with a semicircular dial + digital dial type, that is, the dial has both a semicircular pointer area and a digital dial area, the identification method includes: For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the rectangular table detection is performed to obtain the digital table. For the digital table, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. For the pointer area, the area of the pointer table is first determined by matrix mapping, and then the Hough transform method of finding a straight line is used to determine the position of the pointer. For the found pointer, the fine pointer reading method is used, that is, the distance method is used to identify the fine pointer reading, and the instrument reading is calculated by the distance relationship between the pointer straight line and the adjacent scale line to obtain the pointer recognition result.
3. The dial recognition method based on image processing according to claim 1, characterized in that: For a dial type that is a thick pointer circular dial with only two scales, the identification method includes: For the acquired dial image, we first perform image graying, image defogging, image filtering, and image enhancement preprocessing algorithms in sequence. Then, we perform circular table detection based on Hough transform. In the obtained circular table, we obtain all the contours of the dial by marking the connected areas, marking the strokes, and marking the contours. The largest contour is the location of the pointer. Finally, we determine whether the current state of the pointer is open or closed by the angle of the pointer.
4. The dial recognition method based on image processing according to claim 1, characterized in that: For a round dial whose dial type is a full dial, the identification method includes: For the acquired dial image, we first perform image graying, image dehazing, image filtering, and image enhancement preprocessing algorithms in sequence. Then, we perform circular dial detection based on Hough transform to detect the circular dial in the image. Then, we use regional contrast, regional spatial relationship, center prior saliency features, and correlation matrix algorithm to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to perform the reading recognition of the pointer on the dial.
5. The dial recognition method based on image processing according to claim 1, characterized in that: For a circular dial with multiple pointers, the identification method includes: For the acquired dial image, firstly, the image grayscale, image defogging, image filtering, and image enhancement preprocessing algorithms are executed in sequence, and then the circular dial is detected to detect the circular dial in the image, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image contains a square dial, firstly, the pointer area is determined by the matrix mapping method, and then the region growing, ROI region extraction, and refinement algorithms are used in sequence to obtain the pointer sub-image, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition.
6. A dial recognition device based on image processing, characterized in that: include: Image processing unit. According to the characteristics of the dial, it is divided into four categories through the image processing unit: (1) semicircular dial + digital dial rectangular dial; (2) A round dial with two thick pointers; (3) A full dial; (4) A multi-pointer dial. For the four different types of dials, different image recognition algorithms are used to identify the pointers or numbers on the dial.
7. The dial recognition device based on image processing according to claim 6, characterized in that: For a rectangular dial with a semicircular dial + digital dial type, that is, the dial has both a semicircular pointer area and a digital dial area, the identification device includes: Rectangular dial recognition unit for semicircular meter + digital meter. For the acquired dial image, firstly, perform image graying, image defogging, image filtering, and image enhancement preprocessing algorithms in sequence, and then perform rectangular meter detection to obtain a digital meter. For the digital meter, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition. For the pointer area, first use matrix mapping to determine the area of the pointer meter, and then use the Hough transform to find the straight line method to determine the pointer position. For the pointer found, the fine pointer reading method is used, that is, the distance method is used to identify the fine pointer reading, and the instrument reading is calculated through the distance relationship from the pointer straight line to the adjacent scale line to obtain the pointer recognition result.
8. The dial recognition device based on image processing according to claim 6, characterized in that: For a dial type of a thick pointer circular dial with only two scales, the identification device includes: The thick pointer circular dial recognition unit with only two scales. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the circular table detection based on Hough transform is performed. In the obtained circular table, all the contours of the dial are obtained by marking the connected area, marking the stroke, and marking the contour. The largest contour is found to be the location of the pointer. Finally, the current state of the pointer is determined by the angle of the pointer.
9. The dial recognition device based on image processing according to claim 6, characterized in that: For a round dial whose dial type is a full dial, the identification device includes: Whole dial recognition unit. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are executed in sequence, and then the circular dial detection based on Hough transform is executed to detect the circular dial in the image, and then the regional contrast, regional spatial relationship, center prior saliency feature, and correlation matrix algorithm are used to obtain the prior saliency map, which is optimized through the undirected graph sorting algorithm to perform the reading recognition of the pointer in the dial.
10. The dial recognition method based on image processing according to claim 6, characterized in that: For a circular dial with multiple pointers, the identification device includes: Multi-pointer recognition unit. For the acquired dial image, firstly, the image graying, image defogging, image filtering, and image enhancement preprocessing algorithms are performed in sequence, and then the circular dial is detected. The circular dial in the image is detected, and the pointer sub-image is obtained by using the region growing, ROI region extraction, and refinement algorithms in sequence, and finally the pointer recognition result is obtained; if the image contains a square dial, the pointer area is first determined by the matrix mapping method, and then the pointer sub-image is obtained by using the region growing, ROI region extraction, and refinement algorithms in sequence, and finally the pointer recognition result is obtained; if the image also contains a digital dial, HOG is used for feature extraction, and SVM is used for classifier training for digital recognition.