Design of Single-cut System for Rounded Bag of Bag Making Machine Based on Machine Vision

Design of Single-cut System for Fillet Bag of Bag Making Machine Based on Machine Vision
ABSTRACT: The work aims to design a single-cut system based on machine vision for rounded bags to solve the problems of low efficiency, large amount of waste and fast mechanical wear of the double-cut method often used in production of rounded bags by bag making machines. A visual hardware platform was built. Based on the Open CV visual library, visual software was written in the visual studio development environment. The camera was set in the flying shooting mode, and the output signal was divided to the PLC through the first traction servo encoder. The PLC conducted hard trigger on the camera through the I/O port by virtue of a high-speed counter. A calibration board was used to calibrate the camera to get the pixel equivalent value. After filtering, binarization, edge detection and other processing, the feature of the rounded corner was extracted to make the template. By traversing the image, key points were added on the center line of the template corners, the template center coordinates can be directly obtained when the template matching was completed. The difference between the center coordinates of the corners and the baseline coordinates was sent to the PLC which controlled the tool-shifting servo to realize single-cutter precise cutting. This method can realize fast and accurate positioning of fillet center in pulling the material, and the current bag-making cycle was used to move the knife to cut the current bag. The actual cutting error was within ±0.1 mm, which met the current production requirement on minimum fillet radius 0.125 round corner bag. The vision system adopting the flying mode has high image processing speed, can match the high-speed bag-making requirements, effectively solve the problems that occur in the traditional production of round-cornered bags, meet the needs of industrial production, and have great practical application value. KEY WORDS: rounded corner bag; machine vision; flying shooting; feature extraction; cropping


Bag-making machine as a common machine of soft packaging machinery equipment, can make various forms of packaging bags, its main use of plastic thermoplastic principle, traction and feeding, hot pressing, hot sealing, cutting and other processing technology, will be printed film into packaging bags, widely used in food packaging, light chemical and other production fields. Bag-making mechanism The coordination level of various technological links and parameters in the bag-making process directly affects the quality of bag-making. After years of development, bag-making machine has made great progress in feeding constant tension control, ironing constant temperature control, film material flow control and other processes. Most of the process problems affecting bag making quality depend on the mechanical structure and control system, but it can not be solved simply by optimizing the mechanical structure and control system in the aspects of positioning, cutting and defect detection.
With the development of computer application and image technology, machine vision technology is gradually applied to packaging production process because it can quickly acquire a large amount of information and is easy to process automatically. An image segmentation method was designed based on maximum entropy threshold, and an image denoising algorithm was designed by using adaptive Gaussian guided image filtering algorithm. Machine vision was applied to automatic detection of drugs in drug packaging production line. The machine vision technology was used to detect the defects of the outer packaging of cigarette strips on the production line. At present, the application of machine vision in bag-making machine is less in China. Bag making machine in the production of rounded bag, because of the membrane material printing error and color sensor search error, resulting in the cutter can not be accurately cut in the center of the rounded corner, the production of rounded bag corner will appear obvious burr. At present, bag making machine usually adopts double cutting mode, cutting knife continuous action 2 times, to avoid burr. This method will cut a section of 2 mm or so of waste, not only waste material, but also aggravate the wear and tear of the cutter and the knife moving mechanism, and even waste may be mixed into the finished bag, resulting in product quality problems. Machine vision system because of its high efficiency, strong anti-interference ability, high positioning accuracy, can be used in the corner bag positioning cutting. The visual platform is built, and the precise positioning of the center of the rounded corner bag is realized through soft trigger and the mode of stop and fixed beat.
On this basis, taking the three-side sealing self-standing zipper bag-making machine as the research object, a new control system of single fillet cutting is designed based on machine vision. The camera is installed on the tool rest, and the first traction servo encoder distributes frequency output signals to PLC, which triggers the camera hard through THE I/O port. Compared with the soft trigger, the time accuracy has been greatly improved. The camera adopts the fly-shooting mode, completes the image shooting and processing in the process of drawing, and can accurately cut the current bag in the current bag-making cycle.

1. Overall scheme design
1.1 Analysis of the production process of rounded bag
As shown in Figure 1, the film material is fed by traction rubber roller. When the color label sensor finds the label, the film material stops, and the hot sealing, pressing, punching and cutting actions are completed at this time. In the production of rounded bag, the knife continuous action 2 times can complete the production of rounded bag. Most bag making machine are using this solution to the rounded bag production, although this method can effectively avoid the rounded corners in the emergence of burr, but will cut a long about 2 mm of the waste, waste materials, mechanical structure of the resulting wear and low efficiency problems that cannot be resolved, in addition, when the membrane material tension instability or a printing error, still can appear burr phenomenon.
1.2 Shooting scheme design
As shown in Figure 2, the camera is mounted on the tool rest through the camera bracket, which enables the camera to move and adjust in x, Y and Z directions, and the tool rest can move with the cutter. The shooting position of the camera is between the first traction roller and the cutter. In order to meet the requirements of high-speed bag-making, the camera adopts flying mode. The first traction servo according to the set trigger position through the encoder frequency division output trigger signal to PLC, PLC through high-speed counter counting, when the trigger position reached the camera connected to the I/O port to give trigger signal for hard trigger. Camera installation is shown in Figure 3.




                            

1.3 Vision System Architecture The single fillet cutting system architecture of bag making machine based on machine vision is shown in Figure 4, which is mainly divided into hardware and software. The three-side sealing self-standing zipper bag-making machine adopts Panasonic FP-XH CC60T PLC as the main control unit, and builds a visual hardware platform on this basis. Add industrial computer, 1.3mpp pixel GigE interface black and white industrial face array camera, 2/3" low distortion manual focus lens, customized 6640 mm×20 mm×24 mm strip backlight, dual-channel digital light controller, display and other equipment. Since the first traction servo encoder sends differential signals, while Panasonic FP-XH C60T PLCC can only receive collector signals, the differential to collector signal converter is added. In the Visual Studdio 2015 development environment based on Open CV computer vision library, using C++ language to write Visual processing software, running on Windows 7 system. After triggered by the camera, the image is sent to the vision software through Ethernet. The vision software processes the image and obtains the center coordinates of rounded corners through filtering, binarization processing, edge processing, feature extraction, template matching and other related operations. The industrial computer uses RSS485 serial communication to send the data to PLC, which controls the tool shifting servo for tool shifting and cutting after transformation. See Figure 5 for the visual hardware platform.

2 Software Design
2.11 Camera calibration
Before the camera works, the light source, exposure time, gain, lens aperture and focal length should be adjusted first, so that the camera can take a clear and stable picture. In machine vision, camera calibration is a basic but very key problem. The precision of camera system parameters obtained by camera calibration technology has a great influence on the success of the whole system. Camera calibration is to find the relationship between the position of a point on the surface of a space object in the world coordinate system and its corresponding point in the image coordinate system [12]. The system is mainly to obtain the X-axis coordinates of the center of the fillet and then get the correction value by one step. Therefore, the calibration of the camera can be completed only by completing the relationship between the pixel value of the picture and the actual size, namely establishing the pixel equivalent. The calibration plate shown in FIG. 6 was selected, and the camera took three patterns, including rectangle, circle and triangle, respectively, to calculate the functional relationship between the pixel points of the pattern and the actual size. The obtained pixel equivalents are shown in Table 1.
As can be seen from Table 1, the pixel equivalents obtained by different graphs are similar, and the camera calibration is relatively accurate. In the practical application process, because the visual positioning accuracy is much higher than the precision of the tool moving screw, the pixel equivalent only needs to be 0.02289 to accurately complete the tool moving cutting.



Fig.4 Vision system architecture

 

Image	Pixel equivalent /(mm·pixel−)
orthogon	0.028 8892
ciircle	0.028 8889
right triangle	0.028 8894

Tab.1 Pixel equivalent    v





 Fig.5 Vision Hardware Platform  

 

2.22 Image pee-processing 
The visual system uses image processing to obtain desired data. Because of the interference of the external environment, the photos taken by the camera generally have noise, distortion and other problems. Before image processing, filtering and noise reduction and binarization are usually used for image pee-processing. The commonly used filtering algorithms include mean filtering, Gaussian filtering, median filtering and bilateral filtering. The original image was filtered accordingly, and the original image and each filtering result were shown in Figure 7-8. Through observation and analysis of the four filtering results of the original image with rounded corners, the four filtering algorithms all weaken the noise information, but the mean filtering and Gaussian filtering blur the edge contour and weaken the effective features of the image. Median filter and bilateral filter protect edge information while reducing noise. Combined with the operation time of each filtering algorithm in Table 2, the median filtering algorithm was finally selected.


                                  Fig.6 Calibration board                                           Fig.7 Original picture with rounded corners



    

a median filter                                                    b gaussian filtering                                          c median filtering                                             d Bilateral filtering

Fig.8 Filtering effect

 

Test time	filtering algorithm	operation time /ms
1	average filtering	5.1
2	Gaussian filter	6.8
3	median filter	0.8
4	bilateral filter	2.9

Tab.2 Comparison of computing time of four filtering algorithms

After filtering and denoising, the image needs binarization processing. The basic idea of image binarization is to set the threshold value to divide the image into two parts, the part of original pixel gray value greater than the threshold value is transformed into 255, the part of original pixel gray value less than the threshold value is transformed into 0, and the whole image presents only pure black and pure white visual effect. The commonly used binarization methods are fixed threshold method and Otsu method. Since the images collected are illuminated by backlight, the background is relatively uniform and the brightness difference is small, the fixed threshold method is used for binarization processing.




Where :(x,y) is the coordinate of the input image pixel; G (x,y) is the gray value of point (x,y). According to Figure 9, the gray values of images differ significantly, so the intermediate value of T is 128. The effect of binarization is shown in Figure 10. 


2.3 Feature Extraction
Feature extraction is the process of extracting useful data information from digital images, which can be represented by isolated points, continuous curves or continuous regions. Image features can be generally divided into four categories: color feature, texture feature, shape feature and spatial relation feature.

               

Fig.9 Grayscale histogram                                 Fig.10 Binary result                                     Fig.11 ROI area                                                                  Fig.12 Feature extraction

Combined with the system analysis, the contour shape of rounded corner is taken as the feature information, and the extraction effect is the most ideal. Before feature extraction, region of interest (ROI) needs to be set, as shown in Figure 11. After selecting the favorable interest region, Canny operator was called to extract the edge contour, save the feature region and generate the template, as shown in Figure 12.

2.4 Adding feature points
The basic purpose of template matching is to quickly identify images and find the center coordinates of rounded corners. When the matching is completed, because the coordinates of the center of the rounded corner cannot be obtained directly, the image needs to be further traversed until the coordinates of the center of the rounded corner are found, which will waste a lot of time and affect the operation efficiency of the system. In order to improve the efficiency of software operation, circles are added directly in the template making process
On the basis of template matching, Homogray mapping is used to obtain the central coordinates of the rounded corners of the current image when template matching is completed. This method can significantly improve the efficiency of the visual system, and the comprehensive image processing time is less than 10 ms. The process of adding feature points is described below.
1) On the basis of edge detection, shi-Tomasi algorithm is called to detect corner points of the image, and the obtained corner coordinates O(x,y) are recorded. The detected corner points can be anywhere in the small arc.
2) The image coordinate system takes the upper right corner as the origin, and the X-axis and Y-axis are downward and left respectively. Starting from the origin of the image, the image is traversed column by column along the Y-axis, and the coordinates S(a,b) of the first point with a gray value greater than 128 are recorded. Because the radius of the small arc is 0.25mm, according to the pixel equivalent, if the difference between A and X is less than 9, point S can be determined as the edge point at the center of the circular corner. 3) With S(A,b) as the origin, the image is traversed upward and downward from the right side to the left side respectively. When the gray value of a pixel is greater than 128, the pixel is the contour edge. If the gray value of the pixel point is less than 128, it is not the contour edge. Using this method, 64 lines of edge coordinates of upper and lower contours are detected longitudinally.
4) Calculate the arithmetic mean value of x axis coordinates of 64 groups of upper and lower contour points, which is the symmetry axis of the rounded corners. The y axis coordinates near S can be used to complete the addition of feature points.


 
Where, K is the X coordinate of the center point of the rounded corner; Ui is the x coordinate of the upper contour point; Di is the x-coordinate of the lower contour point.
3 Positioning and cutting
After the center coordinates of rounded corners are obtained, the correction value of the cutter is calculated according to the reference reference point, and the servo controller controls the cutter servo for positioning and cutting. Cutting process: The camera is mounted on the knife rest and can move with the knife. After camera calibration and pixel equivalent, the camera can work. First, turn off the automatic deviation correction function of the cutter, as shown in Figure 13, and roughly adjust the trigger position C so that the camera can capture the rounded picture. Taking the center line of the camera's visual field as the reference line, fine-tune the trigger position C to make the center of the rounded corner coincide with the center of the visual field. Adjust the cutter position so that the cutter can accurately cut into the center of the rounded corner of the bag. At this point, the center of the rounded corner coincides with the center of the visual field, and the cutter can accurately cut to the center of the rounded corner. The reference point is the center of the visual field A, which completes the search of the reference point. On this basis, open the cutter automatic correction function. In the subsequent photos taken by the camera, if the center of the rounded corner coincides with the center of the field of view A, it means that the cutter is cut exactly in the center of the rounded corner, and the error is 0, and the cutter does not need to be adjusted. If the rounded center spacing D with the vision center, prove that needs to be A knife in the same direction after the corresponding distance D to cut into the center of the rounded accurately, the visual software will rectify data (B and A coordinate difference in the x direction) is sent to PLC, PLC after pixel equivalent transformation based on the actual moving distance control servo moving knife knife for accurate cutting. After the cutter is cut, the camera will also move the same distance with the knife. At this time, the center of the rounded corner B and the center of the visual field A will coincide again, forming A new reference point, and the system will operate stably and continuously.
Normally, the deviation correction value is less than the cutter limit value. In order to avoid the interference of external factors, the obtained deviation correction value is greater than the actual deviation correction value, and the cutter reaches the limit. When the correction value is greater than the cutter limit, the interference error can be identified, and the vision system directly sends the error message code (the system chooses to send 4040). At this time, the cutter will not perform cutting action, and the staff will be reminded to check. The workflow of the system is shown in Figure 14. The real-time operation of the system is shown in Figure 15.









Fig.13 Process of cutting



Fig.14 Flow chart of system                                                       Fig.15 Real-time running diagram of system