CN110728279A - A water meter digital recognition method based on embedded platform machine vision - Google Patents

Abstract

The invention discloses a water meter digital recognition method based on embedded platform machine vision, comprising the following steps: A. obtaining a grayscale image on an openMV platform; B. performing Gaussian filtering and binarization on the image obtained in step A; C. In the preprocessed picture in step B, intercept the digital area; D. For the obtained digital area, use vertical projection and horizontal projection to accurately determine the position of the digital character, and segment the digital character; E, Five Search for the best match for each number character in the pre-made template respectively; F. Determine the number of each digit according to the value of the ordinate of the region returned by template matching; G. Return the recognized number to the server, where this In the invention, the water meter digital recognition method based on embedded platform machine vision performs the image recognition step at the terminal, which reduces the amount of data transmission compared with the method of returning an image to realize digital recognition on the server side.

Description

A water meter digital recognition method based on embedded platform machine vision

technical field

The invention relates to the technical field of PCB manufacturing, in particular to a water meter digital identification method based on embedded platform machine vision.

Background technique

With the development of computer technology and image recognition technology, artificial intelligence technology has gradually been applied in people's production and life. As we all know, more than 80% of human perception of external information is realized through vision. Image is the main carrier of information, so the key to information acquisition is image technology, so it is an inevitable development trend to widely use image information processing technology. The development of image processing technology has achieved remarkable results. By using intelligent tools with image recognition technology, it can not only ensure the safety of work, but also liberate labor from complicated and tedious manual labor, thereby greatly improving production efficiency. .

In order to facilitate the management of urban residents' water use, the urban water supply department installs a mechanical water meter for each household to monitor the water consumption of the residents and collect water fees according to the water consumption. At present, water supply departments or enterprises employ special meter readers to copy water meter readings once a month. Manual meter reading has its natural drawbacks. First of all, for large cities, it is necessary to hire a large number of meter reading personnel, which requires a lot of labor and time costs. Secondly, manual meter reading will inevitably make mistakes, and it is difficult to eliminate the phenomenon of copying errors, and it is also necessary to organize paper data, which is also a time-consuming, labor-intensive and error-prone task. Finally, manual meter reading is not real-time enough, and it is impossible to grasp the water consumption situation of residents or enterprises in time. Therefore, some emergency situations, such as serious water leakage and water theft, cannot be dealt with in time, causing serious losses to the water supply department.

With the development of monitoring technology and computer technology, remote automatic meter reading becomes possible. The camera direct reading water meter is one of the technical solutions, which uses a miniature camera installed on the traditional water meter to regularly shoot the water meter dial and transmit it to the server. Remote automatic meter reading. This solution does not need to replace the original water meter, and the image data can be used as a proof, so it is widely welcomed by the market.

Template matching (Tempte Mactch) method is one of the most commonly used methods in image recognition. It extracts several feature quantities in the image or image area with recognition, and then compares them with the corresponding feature quantities in the template one by one. By calculating the normalized correlation between them, finding the one with the largest correlation indicates the highest similarity, and the image can be assigned to the corresponding class. Since only digital characters need to be recognized in the present invention, the selection of the template matching recognition algorithm has the characteristics of high efficiency and speed.

Template matching usually establishes a standard template library first. The standard templates in the library are often digital templates and need to be binarized. At the same time, the size of each character template in the library is uniform. Before template matching, it is often necessary to standardize the character image to make it follow Templates are the same size. At present, the method of template matching is usually used to identify common printed characters. The template matching algorithm is to match the standardized digital character image and template characters one by one to obtain the corresponding matching similarity.

Template matching can be used for the recognition of printed characters because its advantages are that when the characters are relatively regular, it has a strong anti-interference ability to the defect and smudges of the character image, and the recognition rate is also high. After the digital character image of the water meter in the present invention undergoes a series of processing, the character characteristics of the digital character image are well preserved and highlighted, and the template matching algorithm is used to identify the digital characters of the water meter, which can efficiently and accurately identify the water meter. Numeric characters.

However, in addition to the normal single-word image recognition, due to the structural characteristics of the water meter, different degrees of half-words often appear during the reading process. As one of the difficulties in water meter reading recognition, half-word recognition has never had a better recognition algorithm.

In order to be able to recognize half-words, the existing research results often use the following methods: firstly, it is necessary to segment the digital area of half-words, and select a larger area to continue the recognition work. In the segmentation stage, the original image is first binarized, and the original image becomes two areas with obvious white intervals. It can be seen that the image contains obvious blank areas. After that, the best segmentation position is obtained by editing the image pixels. After segmentation, the larger area is selected for subsequent identification. During the recognition process, the feature points of the half-word image will be missing to a certain extent, resulting in a decrease in the accuracy of the recognition.

Therefore, the present invention adopts a special template matching method: ten standard digital templates of uniform size are converted into a digital template of a continuous character wheel. According to the positioning, intercept a certain digit in the image obtained by the camera, and match it in the template. If the match is successful, the corresponding position will be returned. Readings can be determined from location information according to a pre-set location class. The advantages of the present invention lie in that the problem of low recognition rate of half-words is better solved, and there is no need to distinguish whole words and half-words for recognition in advance.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a water meter digital identification method based on embedded platform machine vision, to solve the problems raised in the background technology.

To achieve the purpose, the present invention provides the following technical solutions:

A water meter digital recognition method based on embedded platform machine vision, comprising the following steps:

A. Obtain grayscale images on the openMV platform;

B. Gaussian filtering and binarization are performed on the image obtained in step A;

C, in the preprocessed picture in step B, intercept the digital area;

D. For the obtained digital area, use the method of vertical projection and horizontal projection to accurately determine the position of the digital characters, and divide the digital characters;

E. Five digital characters search for the best match in the pre-made template respectively;

F. Determine the number of each digit according to the value of the ordinate of the area returned by template matching;

G. Return the identified numbers to the server.

As a further scheme of the present invention: the step A is specifically: in the module sensor, call the reset(), set_pixformat(), set_framesize() functions to set the parameters of the camera as: grayscale, and the picture size is QQVGA (160*120) , turn off auto gain, turn off auto white balance.

根据具体安装盒内部空间中LED灯源的位置确定 二值化阈值，对图像对象进行二值化处理。As a further scheme of the present invention: the step B is specifically: the acquired image is an image object, calling the gaussian() function in the image module to eliminate the noise in the image, and then calling the kernel_filter() kernel filter function to sharpen the digital edge, where the kernel is

The binarization threshold is determined according to the position of the LED light source in the internal space of the specific installation box, and the image object is binarized.

As a further scheme of the present invention: the step C is specifically: in the preprocessed picture in step B, use the copy() function in the image module to get a suitable ROI to intercept the digital area, and assign it to a brand-new image object .

As a further solution of the present invention: the step E is specifically: taking the five image objects obtained in the step D as "templates" respectively, and searching for an optimal match in the pre-made pictures.

As a further scheme of the present invention: the step E is specifically: according to the value of the ordinate of the region returned by the template matching of the step E, the number of each digit is determined, and combined into a five-digit number.

As a further solution of the present invention: the step G completes digital transmission through GPRS.

Compared with the prior art, the beneficial effects of the present invention are: the digital identification method of the water meter based on the embedded platform machine vision of the present invention carries out the image identification step at the terminal, and the reduction is reduced compared with the method of returning an image to realize the digital identification at the server end. It reduces the amount of data transmission, perfectly solves the problem of half-words, and is easy to implement.

Description of drawings

Figure 1 is a structural diagram of a water meter digital remote identification technology system based on embedded platform machine vision.

FIG. 2 is a flow chart of the program.

Figure 3 is the digital template of the water meter corresponding to the sample image.

FIG. 4 is a schematic diagram of the identified template position.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

Embodiment 1: Please refer to Figures 1-4, in order to achieve the purpose, the present invention provides the following technical solutions:

A water meter digital recognition method based on embedded platform machine vision, comprising the following steps:

A. Obtain grayscale images on the openMV platform.

In the module sensor, call reset(), set_pixformat(), set_framesize() and other functions to set the camera parameters as: grayscale, picture size as QQVGA(160*120), turn off automatic gain, and turn off automatic white balance.

B. Perform Gaussian filtering and binarization on the image obtained in step A:

The image obtained in step A is an image object, and the gaussian() function in the image module is called to eliminate the noise in the image. Then, call the kernel_filter() kernel filter function to sharpen the digital edges, where the kernel is . The binarization threshold is determined according to the position of the LED light source in the internal space of the specific installation box, and the image object is binarized.

C. In the preprocessed image in step B, use the copy() function in the image module to take a suitable ROI to intercept the digital area and assign it to a brand new image object.

D. For the obtained digital area, use vertical projection and horizontal projection to accurately determine the position of the digital characters, and divide the digital characters.

E. Five precise digital image objects are obtained in step D. Take these five image objects as "templates" respectively, and search for the best match in the pre-made pictures (see Figure 4).

F. Determine the number of each digit according to the value of the regional ordinates returned by the template matching in step E, and combine them into a five-digit number.

G. Send the recognized value back to the server.

Embodiment 2, on the basis of Embodiment 1, step G completes digital transmission through GPRS, the GPRS signal transmission speed is fast, the transmission is stable, and the use is convenient.

It will be apparent to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, but that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, the embodiments are to be regarded in all respects as illustrative and not restrictive, and the scope of the invention is to be defined by the appended claims rather than the foregoing description, which are therefore intended to fall within the scope of the claims. All changes within the meaning and scope of the equivalents of , are included in the present invention. Any reference signs in the claims shall not be construed as limiting the involved claim.

In addition, it should be understood that although this specification is described in terms of embodiments, not each embodiment only includes an independent technical solution, and this description in the specification is only for the sake of clarity, and those skilled in the art should take the specification as a whole , the technical solutions in each embodiment can also be appropriately combined to form other implementations that can be understood by those skilled in the art.

Claims (7)

1. a water meter digital identification method based on embedded platform machine vision, is characterized in that, comprises the following steps: A. Obtain grayscale images on the openMV platform; B. Gaussian filtering and binarization are performed on the image obtained in step A; C, in the preprocessed picture in step B, intercept the digital area; D. For the obtained digital area, use the method of vertical projection and horizontal projection to accurately determine the position of the digital characters, and divide the digital characters; E. Five digital characters search for the best match in the pre-made template respectively; F. Determine the number of each digit according to the value of the ordinate of the area returned by template matching; G. Return the identified numbers to the server. 2. a kind of water meter digital identification method based on embedded platform machine vision according to claim 1, is characterized in that, described step A is specifically: in module sensor, call reset(), set_pixformat(), set_framesize( ) function to set the parameters of the camera as: grayscale, the picture size is QQVGA (160*120), the automatic gain is turned off, and the automatic white balance is turned off. 3. a kind of water meter digital identification method based on embedded platform machine vision according to claim 1, is characterized in that, described step B is specifically: the image acquired is an image object, calls the gaussian () in the image module The function removes the noise in the image, and then calls the kernel_filter() kernel filter function to sharpen the digital edge, where the kernel is

The binarization threshold is determined according to the position of the LED light source in the internal space of the specific installation box, and the image object is binarized. 4. a kind of water meter digital identification method based on embedded platform machine vision according to claim 1, is characterized in that, described step C is specifically: in the picture that is preprocessed in step B, utilize copy in image module The () function takes the appropriate ROI to intercept the digital area and assigns it to a brand new image object. 5. a kind of water meter digital identification method based on embedded platform machine vision according to claim 1, is characterized in that, described step E is specifically: the five image objects that step D obtains are respectively used as " template ", in Search for the best match among pre-made images. 6. a kind of water meter digital identification method based on embedded platform machine vision according to any one of claim 1-5, it is characterized in that, described step E is specifically: according to the template matching of step E the area ordinate of return The value determines the digit of each digit and combines it into a five-digit number. 7. A kind of water meter digital identification method based on embedded platform machine vision according to claim 1, is characterized in that, described step G completes digital transmission through GPRS.

Images