Song Huaibo Lu Changhou 1 Wang Fuchun 2 (1. School of Mechanical Engineering, Shandong University, Jinan 250061, China; 2 Department of Mechatronics and Traffic Engineering, Guilin University of Electronic Technology, Guilin 541004) After the production of matched image arrays and templates, and then identification. Since the algorithm only performs small-scale matching on the region where characters exist, the process of searching for irrelevant data information on the entire image is omitted, and the running time of the algorithm is significantly reduced compared with the traditional template matching algorithm, and the image matching immediacy can be satisfied. Need. Template matching is one of the most primitive and basic pattern recognition methods. It is basically a statistical recognition method. The match between each template and the unknown sample depends on whether each unit on the template matches the corresponding unit on the sample. The template matching algorithm has high precision, can solve the problem of stroke breakage, and has strong anti-interference ability. However, template matching still has too long matching time and is not suitable for use in industrial production. How to improve the matching speed of the algorithm is a key issue. Many improved algorithms have been proposed. Cui Zheng 2 proposed two improved template matching algorithms: template weighting based template matching algorithm and feature block based template matching algorithm. Left force uses quadratic matching error algorithm, first coarse matching, and interlaced interlaced data of template That is, 1/4 of the template data is interlaced and interlaced on the searched image, and then the exact match, that is, the search and matching in the neighborhood of the minimum matching error minimum point, the final result is obtained. Then the result of one match is ZZ2. After all the images are matched, the smallest one is found to be the Wo fruit. It can be seen that the calculation amount of template matching is amazing. Each time the matching is done, mXn sub-subtraction is performed. The template matching pattern is mXn-squared, mXn-1 addition, and the whole image is matched (Width-m+1X( Height-n+1) times. Also passed. It can be seen that the dot block composed of noise points is relatively small, and can be deleted by using the morphological method by setting the number field of the point block 'value T. In MATLAB, The method can be implemented by the bwaieaopen (BW, T) function, where BW represents the image to be processed and T is the pixel number field value. After processing by this function, all the point blocks in the image smaller than the domain value are removed. 2.1.2 Character segmentation and storage In order to avoid misclassification during character segmentation, redundant information due to scratches and the like needs to be removed before segmentation, and the image is expanded according to the morphological principle, so that it is smaller than the structural element. Black spots will be further corroded and eliminated. The key is the choice of structural elements. If the structural element selection is too small, there will still be interference information, and complete structural information cannot be obtained. If the structural element is too large, the useful structural information will be eliminated, resulting in the loss of useful structural information. . Due to the large spacing between the characters, the projection method can be used for segmentation. That is, the vertical projection of the character string is performed, and the non-background point pixels of each column of the image are counted, thereby estimating the gap between the adjacent two characters. The projection analysis is the basic method for dividing many characters. The corresponding function is provided in MATLAB, but to simplify the algorithm, use the find function to find the starting position of the character. The specific method is: 1) Use the find function to get the column information of the character. 2) Arrange the resulting columns in ascending order. 3) Considering that the character may be broken, set the field value of the character interval. As long as the interval between two adjacent columns is larger than the set field value, the column can be regarded as the left and right boundary of the character. In the same way, the find function is used to obtain the line information of the character image, and the upper and lower boundaries of each character can be obtained by combining the obtained column information. 4) A set of arrays of divided images is created using the rank information of the obtained characters. The resulting image is extended one pixel outward during the production process to facilitate template matching. The resulting segmented images are stored one by one in an array of cells provided by MATLAB. Cell arrays are a type of data unique to MATLAB. Each element in an array can be a different data type and can be a different dimension. 2.2 The production of template template image template process requires manual intervention. The acquisition process is as follows: 1) Using the result of image processing, confirm the number of digits in the logo image, set to L1;) Determine the number of digits in the image up and down, Left and right pixel coordinates; 3) segment the characters, create an array of cells, and store the segmented characters in the cell array; 4) extract the corresponding digital sub-images of the digital image to the corresponding positions of the cell array one by one to obtain a template; 5) Make a character comparison table according to the order in which the characters are stored. An array of template cells obtained as described above is shown. 2.3 The template matching template matching method of the sign image still uses the previous matching result, the difference is that the matched object is no longer the whole image, but an array composed of the matched objects. Each object of the array is obtained by expanding a pixel on the basis of the obtained character region. In order to further increase the matching speed, the matching process is added when matching, and the template with the difference is not matched with the image to be matched. Matching is performed, and the matching process is as follows: 1) Each matched object is extracted as an identification object in order. 2) Match the obtained template with the matched object. Calculate the height and width of the template. If the height and width of the template are smaller than the value of the matched object, skip the match and select the next template to match. Set the rejection domain value Trefuse. When the matching result is greater than the domain value, the matching is considered successful, and the closest value is selected as the matching result. The character comparison table is searched according to the needs of the algorithm, and the result is displayed. Otherwise, the match is considered to have failed. Assigning a value to the image for a particular character indicates that the recognition failed. 2.4 results of the first test through computer programming, using MATLAB's numerous function libraries and powerful matrix computing capabilities, using the above algorithm, achieved a good recognition effect, the second character "9" recognition error, the remaining characters can be correctly identified The pre-filtering signal with a recognition rate of 92.86% is subjected to software filtering, and the noise interference is effectively suppressed. The processed signal is as shown. It can be seen that the effect is very obvious after being processed by software filtering. 2 Liquid level parameter query and display In order to make the collected data intuitive and visible, in the design, the dynamic real-time curve is used by the PC to display the change of the liquid level parameter. The graph visually reflects the trend of changes in single or several parameters over a certain period of time. The real-time curve reflects the current value and trend of the monitoring parameters, which is a visual description of the current working state. Dynamic Curve Flow Design Since the PC continuously receives data from the data acquisition system, each time the screen displays the curve drawn by the N data at the most recent time. The data processing algorithm adopted by the system is: first define an array, each time the new data comes, move all the data in the array back one bit, and then insert the new data into the position with the subscript 0, when the data is full. After that, the last data is automatically "squeezed out" by the new data. This always guarantees that the saved N data is up-to-date, and its real-time curve always reflects the correspondence between the output signal and the number of sensor sections. The real-time curve display interface real-time curve interface is as shown. The liquid level sensing signal is scrolled on the screen, which is a series of triangular waves corresponding to the number of sensor sections. According to the triangular wave signal and the number of sensor sections, the liquid level height and volume can be calculated, and the liquid level height change is displayed in real time in the window level dynamic map of the window. The dynamic ratio map of the tank level is displayed on the left side of the window. The left side is the LH2 tank, the right side is the LOX tank, and the right side of the window shows the real-time curve. The medium-solid triangle wave represents the LH2 dotted triangle wave representing the LOX. And the query history curve can more intuitively reflect the trend of one parameter or several parameters in a certain period of time in the past. By analyzing the historical curve, the staff can easily judge the stability of the process, whether the process parameters are set properly, and the cause of the failure. Since all the data is recorded in the database, according to the set query conditions, the corresponding data record set can be obtained through the SQL statement, and displayed in the form of a historical curve on the interface. 3 In the conclusion design, the liquid level data of the data acquisition system is transmitted to the PC in real time by the serial port, and then the data processing is completed in the PC, including software filtering, data storage, numerical calculation, curve display and the like. The system enables online real-time detection of tank (tank) level sensor segments with intuitiveness and visibility. Shaoxing City Shengfeng Import And Export Co.,Ltd. , https://www.sunfeelfabric.com
August 19, 2022