Cigarette warehouse cigarette visual image acquisition and detection device based on heating cigarette packaging machine type

Abstract

The invention provides a cigarette warehouse cigarette visual image acquisition and detection device based on a heating cigarette packaging machine type, which belongs to the technical field of image acquisition and detection, is arranged in a lower cigarette warehouse area of a heating cigarette packaging machine, provides a working phase through an incremental encoder, and adopts an advanced machine vision technology. Further detection and elimination of the loose-end cigarettes are added, meanwhile, relevant data statistics is carried out, normal packaging of other cigarettes is not affected, cigarette production consumption is greatly reduced, the single-shift yield is improved, the working intensity of workers is reduced, and the scientific management concept of quality improvement, consumption reduction and yield increase is met.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image acquisition and detection, and particularly relates to a visual image acquisition and detection device for cigarette in a tobacco warehouse based on a heating cigarette packaging machine. BACKGROUND

[0002] At present, as mentioned in the Chinese patent with the patent publication number "CN113712249A", the packaging equipment, that is, the heating cigarette packaging machine, belongs to one of the packaging machine types and is applied in the production link of new tobacco. In the packaging process of new tobacco, the long-existing quality problem of empty head of cigarettes often occurs. The main solution to this quality problem is to detect through a group of empty head detectors installed at the small package mold box. When the detector detects that there is one or more empty head cigarettes in the mold box, the 20 cigarettes in the mold box will be completely removed from the production line. The removed cigarettes, regardless of good or bad, are treated as waste cigarettes, and the packaging accessories of the cigarette package are also removed. This detection and treatment method causes great waste to the cigarette production and does not conform to the current production management concept of improving quality and reducing consumption. SUMMARY

[0003] The purpose of the application is to provide a visual image acquisition and detection device for cigarettes in a tobacco warehouse based on a heating cigarette packaging machine, which effectively avoids the defects of the detection and treatment method in the prior art, which causes great waste to the cigarette production and does not conform to the current production management concept of improving quality and reducing consumption.

[0004] The application uses the following technical solutions.

[0005] A visual image acquisition and detection device for cigarettes in a tobacco warehouse based on a heating cigarette packaging machine, comprising:

[0006] A mechanical structure and a detection software, comprising:

[0007] A visual probe, comprising four sets of LED light sources and four sets of high-speed cameras;

[0008] An encoder, installed on the moving end of a channel motor connected with a controller, serving as a control phase of each channel motor, and connected with the controller as an IO control board to control the working time of each visual probe;

[0009] An alarm mechanism, installed above a display connected with the controller and connected with a computer;

[0010] A control cabinet, comprising a computer connected with the controller and an IO control board;

[0011] The detection software runs on the controller.

[0012] Further, the detection software adopts Japanese FAST, Canadian CORECO algorithm library, German HALCON algorithm library, which contains a large number of basic functions of image processing; the detection software performs image recognition on the detected part of the cigarette, that is, firstly, the standard image of qualified cigarettes is stored, the image of the cigarette is detected in real time in the production process, and is compared with the stored standard image of qualified cigarettes, if the histogram, similarity and image position coordinates used for comparison are within the set tolerance, and the rechecking function is qualified, it indicates that the tobacco side of the cigarette is qualified, otherwise, it is unqualified, and is removed by the original machine.

[0013] Further, the detection software generates an artificial intelligence data model through learning of a set amount of qualified tobacco ends and filter rod ends after training by the controller, and the artificial intelligence data model can be used to judge and identify common defects such as empty head, reverse cigarette, and foreign matter in any part of the cigarette.

[0014] Further, the detection software firstly collects a standard image and generates a template based on the NCC algorithm of integral area; all defect images are collected for training, and the allowed defect tolerance is set.

[0015] Further, the NCC algorithm based on integral area includes:

[0016] The integral image formula of the sum calculated from the top to the bottom and from the left to the right of the standard image of the cigarette is as follows:

[0017] ;

[0018] Wherein represents the sum of the gray values of all pixel points from the top left corner (0, 0) of the image to the current pixel point (x, y), that is, the value of the integral image at (x, y), represents the gray value of the pixel point at (x, y) of the image, represents the value of the integral image at (x-1, y), that is, the sum of the gray values of all pixel points from (0, 0) to (x-1, y), represents the value of the integral image at (x, y-1), that is, the sum of the gray values of all pixel points from (0, 0) to (x, y-1), represents the value of the integral image at (x-1, y-1), that is, the sum of the gray values of all pixel points from (0, 0) to (x-1, y-1);

[0019] Then, the NCC algorithm is executed, and the calculation formula of the NCC algorithm is as follows:

[0020] ;

[0021] Wherein denotes a normalized cross-correlation coefficient, which is used to measure the similarity between the template image and the image to be detected, denotes the size of the calculation window, where m is the size of the height direction of the window, n is the size of the width direction of the window, and (x, y) ∈ m × n, that is, the current pixel point (x, y) is located in the window range, denotes the pixel value of the template image at (x+i, y+j), denotes the pixel value of the image to be detected at (x+i, y+j), denotes the mean value of the template image in any window, denotes the mean value of the image to be detected in any window;

[0022] Then, the template image mean value calculation is performed, and the calculation formula is:

[0023] .

[0024] Further, the LED light source comprises an LED light source strip and an LED light source fixing support for supporting the LED light source strip.

[0025] Further, a hollow area is reserved below the cigarette unloading device and behind the cigarette warehouse, and a visual probe is installed in the hollow area.

[0026] Further, the heating cigarette machine orderly conveys the cigarettes through 20 single-channel passages to a detection position, and the visual probe can shoot 2 cigarettes in the same passage at the detection position.

[0027] Further, the control cabinet is independently installed beside the machine table of the heating cigarette machine.

[0028] Further, the display is independently hung at a position convenient for the operator to view.

[0029] Further, based on the working of the cigarette warehouse cigarette visual image acquisition and detection device of the heating cigarette packaging machine, after the cigarettes enter 20 small passages after being divided, the controller judges that the cigarettes have been positioned according to the encoder signal, controls the LED light source to flash, sends a camera pulse to the high-speed camera, and the high-speed camera transmits the acquired cigarette image to the detection software running on the controller through the USB3.0 port, the detection software analyzes and processes in real time through the AI deep learning algorithm, determines whether the cigarette is qualified, finally, the controller sends the good or bad signal to the computer according to the processing result of the cigarette image, and the computer generates subsequent actions according to the good or bad signal, if the current cigarette is a bad cigarette, that is, the computer receives a bad signal, the air nozzle is removed or the alarm mechanism is controlled to alarm, if it is a good cigarette, that is, the computer receives a good signal, the cigarette is transported to the next production process.

[0030] The beneficial effects of the present application are that, compared with the prior art, the technical effects of the present application include:

[0031] The present application sets up a cigarette magazine area under a heating cigarette packaging machine to provide a working phase with an incremental encoder, uses advanced machine vision technology, increases further detection and rejection of empty-end cigarettes, simultaneously carries out relevant data statistics, does not affect normal packaging of other cigarettes, greatly reduces cigarette production consumption, improves single-shift output, reduces personnel work intensity, and meets the scientific management concept of improving quality, reducing consumption and increasing production. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a cigarette magazine element position diagram of the heating cigarette machine described in the present application;

[0033] Figure 2 is a schematic diagram of the visual probe described in the present application;

[0034] Figure 3 is a whole schematic diagram of the magazine cigarette visual image acquisition and detection device based on the heating cigarette packaging machine type described in the present application;

[0035] Figure 4 is a principle diagram of the continuous diffuse reflection illumination described in the present application;

[0036] Figure 5 is a flowchart of the detection software described in the present application. DETAILED DESCRIPTION

[0037] In the process of new tobacco cigarette packaging, there is often a long-standing quality problem of empty-end cigarettes. The main solution to this quality problem is to detect through a group of empty-end detectors installed at the small package mold box. When the detector detects one or more empty-end cigarettes in the mold box, the 20 cigarettes in the mold box will be completely rejected from the production line. The rejected cigarettes, regardless of good or bad, are treated as waste cigarettes, and the packaging accessories of the cigarette package are also rejected. This detection and treatment method causes great waste to cigarette production and does not meet the current production management concept of improving quality and reducing consumption.

[0038] The present application is in the detection mode of "high-speed CCD industrial camera + PC image processing software" (PC-BASE). This mode has a qualitative improvement in performance indicators, ease of use and maintenance, and upgrade and expansion, and can effectively meet the detection needs of cigarette factory customers. It can achieve online detection of each cigarette. When occasional defective cigarettes occur, the device automatically tracks the defective cigarettes and gives a precise early warning at the appropriate position. When consecutive defective cigarettes occur and reach the set value of consecutive defective number set by the quality department, the device immediately sends a stop signal to the tipping machine and gives an audible and visual alarm to prompt the machine operator to handle it in time.

[0039] The device is arranged in the lower tobacco warehouse area of a heating cigarette packaging machine, provides a working phase by using an incremental encoder, uses advanced machine vision technology, further detects and rejects empty-end cigarettes, simultaneously performs relevant data statistics, does not affect the normal packaging of other cigarettes, greatly reduces cigarette production consumption, improves single-shift output, reduces personnel work intensity, and meets the scientific management concept of improving quality, reducing consumption and increasing output.

[0040] As far as the current domestic situation is concerned, the detection mode or product is still in the initial stage, and there are few similar products on the market. The product has wide development space and will be welcomed by the majority of cigarette factory customers.

[0041] In order to make the purpose, technical scheme and advantages of the present application more clear, the technical scheme of the present application will be expressed clearly and completely in combination with the drawings in the embodiments of the present application. The embodiments expressed in the present application are only partial embodiments of the present application, but not all embodiments. According to the spirit of the present application, other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0042] The tobacco warehouse cigarette visual image acquisition and detection device based on a heating cigarette packaging machine type comprises:

[0043] At present, the common way to detect the quality defects of empty-end cigarettes on the market is machine vision detection, and the detection modes adopted by various manufacturers are roughly divided into the following three types:

[0044] ① Intelligent camera + PC

[0045] The algorithm in the intelligent camera + PC mode is packaged into a fixed module, and the scalability is poor. Different series of cameras need to be configured according to different detection speeds, and the project cost is high.

[0046] ② Camera + image acquisition card + PC

[0047] The camera + image acquisition card + PC mode uses an image acquisition card to compress and output images to reduce hardware configuration, and the point-to-point connection is limited in the number of connected devices, and the stability of the device is low. The looseness of the card slot and the interface will affect the system stability.

[0048] ③ USB camera + PC

[0049] The USB camera + PC mode transmits the images collected by the camera to the industrial computer through the USB3.0 interface, and has stable function, fast speed, moderate cost, customizable hardware configuration, good scalability, realizes the detection function by using the PC-Base system. It can be designed specifically, flexible and variable to meet customer needs. At present, this detection mode is generally used.

[0050] The device of the application comprises a controller, a visual probe, an encoder, an alarm mechanism, a display, and specifically comprises:

[0051] A mechanical structure and detection software, which comprises:

[0052] The visual probe 5 comprises four sets of LED light sources 3, four sets of high-speed cameras 4, and other mechanical accessories, and detects the quality defects of the cigarettes in the four channels of the first lower cigarette warehouse, the second lower cigarette warehouse, the third lower cigarette warehouse, and the fourth lower cigarette warehouse, respectively.

[0053] The encoder is installed on the moving end of the channel motor connected with the controller, and is used to control the working phase of each channel motor.

[0054] The alarm mechanism is installed above the display 7 connected with the controller and connected with the computer, and sounds when unqualified cigarettes appear, prompting the operator to further process the waste.

[0055] The control cabinet is the core of the whole system, and comprises one set of computer connected with the controller and one set of IO control board (controller).

[0056] The detection software runs on the controller, and is programmed with C language based on the modular programming idea in the software design of the controller.

[0057] The software of the visual image acquisition and detection device of the cigarette warehouse of the heated cigarette model mainly comprises the host computer interface of the detection software, the detection software tool, the communication interface, and the CCD image acquisition component communication.

[0058] The host computer interface is a key human-machine interface design, and mainly comprises multiple sub-interfaces, such as a brand template selection interface, a working interface, a debugging interface, and a system setting interface.

[0059] In the preferred but non-limiting embodiments of the present application, the detection software adopts the mature Japanese FAST, Canadian CORECO algorithm library, and German HALCON algorithm library, which contains a large number of basic image processing functions and comprehensively applies deep algorithms; the detection software is the latest version and can be upgraded free of charge. The detection software performs image recognition on the detected parts of the cigarette, that is, first, the standard image of the qualified cigarette is stored, the image of the cigarette is detected in real time during production, and compared with the stored standard image of the qualified cigarette. If the histogram, similarity (FIT), image position coordinates (X, Y) used for comparison are within the set tolerance, and the re-inspection function is qualified, it indicates that the tobacco side of the cigarette is qualified, otherwise it is unqualified, and is rejected by the original machine.

[0060] In the preferred but non-limiting embodiments of the present application, as shown in the figure, Figure 5 The detection software is based on AI deep learning architecture. It has strong processing capability and can adapt to the processing needs of complex algorithms, and has the functions of intelligent AI deep learning and traditional detection dual mode. The detection software learns from a set amount of qualified tobacco end and plug end, generates an artificial intelligence data model after controller training, and uses the artificial intelligence data model to accurately judge and identify common defects such as empty head, reverse cigarette, and foreign matter in any part of the cigarette.

[0061] The fine artificial intelligence data model sets the AI deep learning network model for detecting the target of the cigarette, which only needs a small amount of defect samples for training, and no longer needs manual image configuration and training.

[0062] At the same time, the software system also supports traditional detection mode, and common algorithm tools can be selected for use, such as image search, gray detection, and straight line search algorithm tools. Customers can flexibly select and match according to their needs.

[0063] In addition, the detection software adopts a database mode and comprehensively stores related data, which can be accessed by the workshop MES system at any time. The data is comprehensive, the data communication is stable and reliable, and it does not affect normal detection.

[0064] In order to meet the requirements of intelligent manufacturing and lean production improvement of cigarette enterprises, the artificial intelligence cigarette empty head detection system of the packaging machine group adopts the "AI deep learning" algorithm in the field of artificial intelligence and embeds the AI deep learning algorithm in the GPU of the system for running.

[0065] AI deep learning is a branch of machine learning, which is an algorithm based on artificial neural network architecture for data representation learning. So far, several AI deep learning frameworks have been applied in computer vision, speech recognition, natural language processing, audio recognition and bioinformatics, and have achieved excellent results.

[0066] The image recognition algorithm based on AI deep learning is superior to the traditional image recognition algorithm. The traditional image recognition algorithm is mainly based on low-level feature extraction at the pixel level, and then the extracted features can be classified and recognized. This process cannot meet the task of high-level semantic extraction. Deep neural network well simulates the human visual processing process. In the early stage of neural network model training, the model automatically extracts the most information features from the original image, and then maps to the next layer. Through continuous iteration of this process, high-level semantic extraction is realized, and the recognition result is given.

[0067] Comparison and analysis of traditional detection system and AI deep learning software system:

[0068] 1. The traditional machine vision detection is a threshold plus detection frame mode, which cannot well generalize the change of cigarette blank. A set of algorithm suitable for slim cigarette, medium cigarette or standard cigarette needs to be redesigned to perform well. The AI deep learning system is a neural network algorithm with very good generalization ability. For the change of cigarette, only the defect picture of the target cigarette needs to be input for self-learning detection, without the need to reprogram the algorithm to achieve 98% accuracy. With the accumulation of time, the accuracy of AI deep learning algorithm is higher, which can approach 100%.

[0069] For example, according to the sampling survey of cigarette brand production process, the detection of cigarette and the resulting cigarette blank are compared with the standard size of cigarette. It is found that the recognition accuracy of the conventional detection system is low, and under the strict judgment standard, it causes more misidentification and causes more waste of raw materials. Therefore, it is necessary to improve the detection and recognition accuracy of the detection software for cigarettes, and reduce the misidentification rate.

[0070] 2. For different standard cigarette brands or medium cigarette, slim cigarette, the traditional cigarette blank visual detection system based on fixed programmed algorithm is difficult to identify. AI deep learning detection system is particularly good at solving visual applications that are difficult to use rule-based programs. It has very good effect on processing cluttered background and detecting various changes in workpiece appearance. Only new cigarette image data needs to be pre-trained on site, without the need to redesign the core neural network.

[0071] For example: according to the survey, due to the design of the cigarette stick mouth rod is more exquisite than the previous product, in order to increase the suction group, with various shapes of hollow mouth rod, the identification accuracy of the detection system is put forward higher requirements. In response to many new products, the accuracy cannot meet the requirements. Therefore, it is necessary to improve the identification ability of the detection system for complex cigarette hollow.

[0072] 3. The traditional visual detection system has high requirements for the positioning detection of cigarette. The swing and jitter of standard cigarettes in the production process can easily cause misjudgment of the traditional detection system, resulting in misjudgment of qualified cigarettes, causing material waste. The principle of the existing threshold type detection is that each detection is an independent process, which only relies on the pre-set threshold value of each block area for judgment. The visual quality detection principle of AI deep learning image recognition is to learn a certain amount of qualified cigarettes and defective cigarettes. After the computer AI deep learning neural network training, the data model of artificial intelligence neural network is generated. The data model can accurately judge and identify the hollow defects of any part. AI deep learning algorithm is based on pixel-level defect detection and judgment. The detection standard is composed of a large number of learned historical data. Each detection will contact the "good" and "bad" judgment of 10000 or even 100000 cigarettes in the past. It is like an experienced technician. Therefore, under the premise of sufficient sample size, it can detect extremely fine differences and achieve extremely high detection accuracy. The difference between the two can be described as:

[0073] Threshold type: "hollow down > 3mm → gray scale exceeds threshold, reject"

[0074] AI deep learning type: "experience tells me that the cigarette hollow has a problem, reject".

[0075] Therefore, the AI deep learning detection system is not limited by the running state of the cigarette, and has excellent adaptability to the jitter and swing of the cigarette. It fundamentally solves the problem of cigarette misjudgment caused by jitter and swing, saves materials and costs, and greatly helps to improve production efficiency and product quality.

[0076] 4. The AI deep learning detection system has the ability to correct the detection standard. After the AI deep learning visual detection forms the training model, it can also be automatically corrected in actual work. The flow of each good cigarette will again strengthen the cognition of "good cigarette", and the rejection of each bad cigarette will also again clarify the judgment of "bad cigarette". The standard of good and bad will be fully remembered and deepened in the neural network. Through countless self-adjustment and correction, the detection accuracy will be continuously improved with the accumulation of use time, so as to achieve infinitely close to 100% correct detection, which is the intelligent advantage that the traditional appearance visual detection system cannot match.

[0077] In the preferred but non-limiting embodiment of the present application, the present application focuses on the in-depth research of the image processing algorithm of the current cigarette industry air hole detection system. The detection software of the present application firstly collects a standard image and generates a template based on the NCC algorithm of integral area; collects all defect images for training and sets the allowed defect tolerance. The detection image is collected by using a double-station CCD sensor, and the image is aligned, fused and light migration corrected. The present application overcomes the detection difficulty caused by reflection, has wide detection range, simple operation, high detection precision and high detection accuracy.

[0078] In the preferred but non-limiting embodiment of the present application, the NCC algorithm based on integral area includes:

[0079] The integral image formula of the sum calculated from the standard image of the cigarette from top to bottom and from left to right is as follows:

[0080] ;

[0081] Wherein represents the sum of the gray values of all pixel points from the upper left corner (0, 0) of the image to the current pixel point (x, y), i.e. the value of the integral image at (x, y), represents the gray value of the pixel point at (x, y) of the image, represents the value of the integral image at (x-1, y), i.e. the sum of the gray values of all pixel points from (0, 0) to (x-1, y), represents the value of the integral image at (x, y-1), i.e. the sum of the gray values of all pixel points from (0, 0) to (x, y-1), represents the value of the integral image at (x-1, y-1), i.e. the sum of the gray values of all pixel points from (0, 0) to (x-1, y-1);

[0082] Then the NCC algorithm is executed, and the calculation formula of the NCC algorithm is:

[0083] ;

[0084] Wherein represents the normalized cross-correlation coefficient, which is used to measure the similarity between the template image and the image to be detected, represents the size of the calculation window, wherein m is the height direction size of the window, n is the width direction size of the window, and (x, y) ∈ m × n, i.e. the current pixel point (x, y) is located within the window, represents the pixel value of the template image at (x+i, y+j), represents the pixel value of the image to be detected at (x+i, y+j), represents the mean value of the template image in any window, This represents the mean of the image to be detected within any window;

[0085] Then, the mean of the template image is calculated using the following formula:

[0086] .

[0087] The NCC formula is used to calculate the integral image. After establishing the sum and square sum of the image to be detected and the template, as well as their cross product, five integral image indices are built in the window below. This completes the entire processing and finally outputs the calculation result.

[0088] In a preferred but non-limiting embodiment of the present invention, the LED light source 3 includes an LED light source strip and mechanical components such as an LED light source fixing bracket for supporting the LED light source strip.

[0089] The light source strip consists of LED high-brightness lighting units and heat sinks, providing a lighting mechanism for the device. It highlights the characteristic points of the cigarette cross-section and effectively suppresses reflections from the splicing paper and the influence of external light, achieving uniform illumination. Considering stability, high brightness, and long lifespan (flicker mode, lifespan ≥ 100,000 hours), the device uses an imported, well-known ES brand 24V DC LED high-brightness cool white light source, with a brightness of 500-600LM, long lifespan, and low light decay. It also features a PWM soft constant-on mode, eliminating flicker and not interfering with the normal work of on-site personnel. The image is clear, without light spots, and the brightness is adjustable.

[0090] Simultaneous time-sharing exposure control illuminates the cigarette from different angles, avoiding interference from light sources of different high-speed cameras. It is highly adaptable to various brands of cigarettes and can meet the image inspection requirements of various cigarettes.

[0091] General lighting typically employs strip or spot lighting. Strip lights are a common general lighting method, easily mounted on lenses, and provide sufficient illumination to diffuse surfaces. This device uses multiple light source strips to form an arc-shaped diffuse lighting effect, creating strong parallel light to prevent dark or overexposed areas on the cigarette, facilitating processing by the detection software.

[0092] Continuous diffuse lighting is used on reflective surfaces or surfaces with complex angles, such as... Figure 4 As shown. Continuous diffuse illumination uses a hemispherical, uniform illumination to reduce shadows and specular reflections. This illumination method is very useful for illuminating fully assembled circuit boards. This machine vision light source can achieve uniform illumination over a 170° solid angle.

[0093] In a preferred but non-limiting embodiment of the present invention, the location diagram of the cigarette holder element in the heated cigarette machine is as follows: Figure 1 and Figure 2As shown, in order to realize online detection of single cigarette 1, a hollow area is reserved below the cigarette unloading equipment and behind the cigarette warehouse 2, and a visual probe 5 is installed in the hollow area. In the tobacco production system, the "cigarette warehouse" of the cigarette unloading equipment is the core cache device in the automatic material circulation system of the cigarette making workshop, mainly used for temporarily storing the cigarette semi-finished products produced by the rolling process and connecting the subsequent packaging process. As a temporary storage unit between the rolling machine and the packaging machine, the cigarette warehouse effectively balances the production rate difference between the front and rear processes by caching cigarettes, avoids interruption of the rolling process due to temporary shutdown of the packaging machine, and significantly improves the continuity of the production line. The cigarette warehouse is also the lower cigarette warehouse (main machine).

[0094] In the preferred but non-limiting embodiment of the application, the heating cigarette machine orderly delivers the cigarettes through 20 single-thread channels to the detection position, and the visual probe can take pictures of 2 cigarettes next to each other in the same channel at the detection position, a total of 40 cigarettes (2*20) can be taken pictures at the same time.

[0095] If the detection and identification are good cigarettes, the controller controls the main machine connected thereto to stop and the alarm mechanism to start, informing the operator that the current cigarette has defects and needs to be processed as waste. The cigarettes in a single channel are arranged next to each other, and the visual probe only needs to take pictures of the field of view of the upper and lower two cigarettes.

[0096] In the preferred but non-limiting embodiment of the application, the control cabinet 6 is independently installed beside the machine table of the heating cigarette machine, which is convenient for the user to operate and check the running dynamics of the device.

[0097] In the preferred but non-limiting embodiment of the application, the display installation position can be customized according to different machine types, such as Figure 3 As shown, the display 7 can be independently hung at a position convenient for the operator to check.

[0098] In order to meet the production speed of cigarettes and the defect detection rate of the measured object, the high-speed camera adopts a 1.6 million pixel USB color camera with a frame rate of 227 fps. The exposure timing of the high-speed camera is controlled by the TRIG port, and the camera is exposed once every time the TRIG port jumps. The TRIG port of the camera has a common positive line and a signal line. In actual application, the signal line is easily disturbed by electromagnetic and static electricity, causing the camera to take pictures incorrectly. Therefore, the output signal of the controller needs to be first optically isolated, then filtered by a constant current filter, and finally connected to the TRIG input end of the camera. This effectively prevents incorrect shooting.

[0099] The high-speed camera adopts a global exposure Sony IMX273 CMOS photosensitive chip, transmits image data through a USB3.0 data interface, integrates an I / O (GPIO) interface, provides a cable locking device, can stably work in various harsh environments, and is a high-reliability, high-performance industrial digital camera product.

[0100] In a preferred but non-limiting embodiment of the present application, based on the visual image acquisition detection device of the heating cigarette packaging machine type of tobacco warehouse, when the cigarette enters the 20 small channels after being separated, the cigarette separation mechanism is normally operated. Then, the controller determines that the cigarette has been positioned according to the encoder signal, controls the LED light source to flash, sends a camera pulse to the high-speed camera, and the high-speed camera transmits the collected cigarette image to the detection software running on the controller through the USB3.0 port. The detection software analyzes and processes in real time through AI deep learning algorithm, determines whether the cigarette is qualified, and finally the controller sends good or bad signals to the computer according to the processing result of the cigarette image. The computer performs subsequent actions according to the good or bad signal. If the current cigarette is a bad cigarette, that is, the computer receives a bad signal, the air nozzle is removed or the alarm mechanism is controlled to alarm. If it is a good cigarette, that is, the computer receives a good signal, the cigarette is transported to the next production process.

[0101] The beneficial effects of the present application are that, compared with the prior art, the technical effects of the present application include:

[0102] The present application is installed in the lower tobacco warehouse area of the heating cigarette packaging machine, provides a working phase with an incremental encoder, uses advanced machine vision technology, increases further detection and rejection of empty head cigarettes, and performs related data statistics without affecting the normal packaging of other cigarettes, greatly reduces cigarette production consumption, improves single shift output, reduces personnel work intensity, and meets the scientific management concept of improving quality, reducing consumption and increasing production.

[0103] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the above examples, those skilled in the art will understand that modifications or equivalent replacements can still be made to the specific embodiments of the present application without departing from the spirit and scope of the present application. Any modification or equivalent replacement should be covered within the protection scope of the claims of the present application.

Claims

1. A kind of based on heating cigarette packaging machine type tobacco warehouse cigarette visual image acquisition detection device, it is characterized in that, It comprises: a mechanical structure and detection software, which comprises: a visual probe, the visual probe comprising four sets of LED light sources and four sets of high-speed cameras; an encoder, which is installed on the moving end of a channel motor connected with a controller, and is used to control the working phase of each channel motor. The encoder is connected with the controller as an IO control board, and the IO control board controls the working time of each visual probe; an alarm mechanism, which is installed above a display connected with the controller and connected with a computer; a control cabinet, which comprises a computer connected with the controller and an IO control board; the detection software runs on the controller.

2. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 1, characterized in that, The detection software adopts Japanese FAST, Canadian CORECO algorithm library and German HALCON algorithm library, which contains a large number of basic image processing functions. The detection software performs image recognition on the detected part of the cigarette, that is, the standard image of the qualified cigarette is stored first, and the image of the cigarette is detected in real time during the production process, and compared with the stored standard image of the qualified cigarette. If the histogram, similarity and image position coordinates used for comparison are within the set tolerance, and the rechecking function is qualified, it means that the tobacco side of the cigarette is qualified, otherwise it is unqualified, and the original machine is rejected.

3. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 2, characterized in that, The detection software learns from a set amount of qualified tobacco end and filter rod end, generates an artificial intelligence data model after training by the controller, and uses the artificial intelligence data model to judge and identify common defects such as empty head, reverse cigarette and foreign matter in any part of the cigarette. The detection software first collects standard images and generates a template based on the NCC algorithm of integral area. All defect images are collected for training, and the allowed defect tolerance is set.

4. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 3, characterized in that, The NCC algorithm based on integral area comprises: The integral image formula of the sum is calculated from the top to the bottom and from the left to the right of the standard image of the cigarette as follows: ； wherein represents the sum of the gray values of all pixels from the top left corner (0, 0) of the image to the current pixel point (x, y), i.e. the value of the integral image at (x, y), represents the gray value of the pixel at (x, y) of the image, represents the value of the integral image at (x-1, y), i.e. the sum of the gray values of all pixels from (0, 0) to (x-1, y), represents the value of the integral image at (x, y-1), i.e. the sum of the gray values of all pixels from (0, 0) to (x, y-1), represents the value of the integral image at (x-1, y-1), i.e. the sum of the gray values of all pixels from (0, 0) to (x-1, y-1); Then the NCC algorithm is executed, and the calculation formula of the NCC algorithm is: ； wherein denotes a normalized cross-correlation coefficient, used to measure the similarity between the template image and the image to be detected, denotes the size of the calculation window, wherein m is the height direction size of the window, n is the width direction size of the window, and (x, y) ∈ m × n, i.e. the current pixel point (x, y) is located in the range of the window, denotes the pixel value of the template image at (x+i, y+j), denotes the pixel value of the image to be detected at (x+i, y+j), denotes the mean value of the template image in any window, denotes the mean value of the image to be detected in any window; Subsequently, the template image mean value calculation is executed, and the calculation formula is: 。 5. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 4, characterized in that, The LED light source comprises an LED light source strip and an LED light source fixing bracket for supporting the LED light source strip.

6. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 5, characterized in that, A hollow area is reserved below the cigarette unloading equipment and behind the cigarette warehouse, and a visual probe is installed in the hollow area.

7. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 6, characterized in that, The heated cigarette machine orderly transports the cigarettes through 20 single-channel passages to the detection position, and the visual probe can shoot 2 cigarettes next to each other up and down in the same passage at the detection position, and a total of 40 cigarettes can be shot at the same time.

8. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 7, characterized in that, The control cabinet is independently installed beside the machine table of the heated cigarette machine.

9. The visual image acquisition detection device for the tobacco library cigarette of the heating cigarette packaging machine type according to claim 8, characterized in that, The display is independently hung at a position convenient for the operator to view.

10. The visual image acquisition detection device for the tobacco warehouse cigarette based on the heating cigarette packaging machine type according to claim 9, characterized in that, Based on the heating cigarette packaging machine type of tobacco warehouse cigarette visual image acquisition detection device when the work, when the cigarette into 20 small channels after the branch, then, the controller according to the encoder signal, judge cigarette has been in place, control LED light flash, send camera pulse to high speed camera, high speed camera through USB3.0 port will collect the cigarette image transmission to the controller running detection software, detection software through AI deep learning algorithm real-time analysis online processing, determine whether the cigarette is qualified, finally, the controller according to the processing result of cigarette image to computer sends good or bad signal, computer according to good or bad signal occurs subsequent action, if the current cigarette is bad cigarette, that is, the computer receives bad signal, through the rejection nozzle or control alarm mechanism alarm, if for good cigarette, that is, the computer receives good signal, the cigarette is transported to the next production process.

Images