Fully automatic tissue culture bottle filling line and control method thereof

Abstract

The present invention proposes a fully automatic tissue culture bottle filling line and a control method thereof. The line covers workstations such as loading, loading, unscrambling, inverting, degreasing, cleaning, filling, capping, unloading, transporting and sterilizing. The robotic arm and the conveyor line work together to realize the automated processing flow of tissue culture bottles. The inverting bottle station uses an image recognition algorithm to accurately detect and process overturned bottles. The visual detection mechanism includes image acquisition, preprocessing, feature extraction and matching steps. The camera parameters are optimized according to the conveying speed and light intensity. The composite filtering and correction algorithm improves the image quality. The multi-feature group vector input model is extracted to discriminate the bottle posture. The control method orderly directs the operation of each workstation to ensure efficient and accurate processing of tissue culture bottles, improve production efficiency and product quality stability, which is of great significance to the automation upgrade of the tissue culture industry, reduces labor costs, and meets the needs of large-scale production.

Description

Technical Field

[0001] The invention relates to the technical field of tissue culture bottle cleaning, in particular to a full-automatic tissue culture bottle filling production line and a control method thereof. Background Art

[0002] In the factory-based tissue culture production of seedlings of orchids and flowers of traditional Chinese medicine, a huge amount of tissue culture bottles are used. In order to reduce costs, glass tissue culture bottles are usually used, which can be reused after cleaning. Our institute has applied for patent CN117502244A to disclose a tissue culture container heating, cleaning and filling production line, which includes a loading station, a pre-warming station, a fat removal station, a tissue culture container cleaning station, a tightening station and a finishing station arranged in sequence; a conveyor chain is arranged between adjacent stations; and a cap delivery station is arranged on one side of the tightening station. By heating the tissue culture bottle at the pre-warming station, the culture matrix at the bottom of the bottle is liquefied, the adhesion of the culture matrix is ​​reduced, the difficulty of cleaning the culture matrix in the tissue culture bottle is reduced, the blockage of the sewer pipe is reduced, and environmental pollution is reduced. The residual culture matrix in the tissue culture bottle after liquefaction is scraped and ground at the fat removal station to improve the cleaning efficiency and fat removal rate. By replacing manual cleaning with machines and without the need to use hydraulic cleaning of culture bottles, the solid culture matrix attached to the culture bottles can be removed, and the production of tissue culture bottles can be directly completed through fine washing and filling stations, which improves work efficiency, reduces manual labor intensity, reduces water resource waste, and reduces environmental pollution.

[0003] However, although patent CN117502244A has achieved full automation to a certain extent, it is still insufficient in terms of the degree of automation. It is unable to process the tissue culture bottles that have fallen in a batch of tissue culture bottles, and requires manual intervention in advance. It is also unable to realize automatic unloading and transportation to the sterilizing pot to wait for sterilization operations.

[0004] Therefore, a new fully automatic tissue culture bottle filling line and a control method thereof are very much needed to solve the current technical problems. Summary of the invention

[0005] The embodiment of the present invention provides a fully automatic tissue culture bottle filling production line and a control method thereof, which aims to solve the problems of insufficient automation and inability to realize bottle inversion processing in the current technology.

[0006] The core technology of the present invention is mainly to upgrade and transform the original assembly line, by adding additional bottle inverting stations and some new equipment, combined with a new set of control methods, to achieve more efficient fully automatic tissue culture bottle filling.

[0007] In a first aspect, the present invention provides a fully automatic tissue culture bottle filling line, comprising:

[0008] The feeding conveyor line is used to input tissue culture bottle frames to be cleaned, and each tissue culture bottle frame is filled with tissue culture bottles;

[0009] The bottle loading station is located at one side of the end of the loading conveyor line. The bottle loading station is provided with a bottle loading robot arm whose working range covers the end of the loading conveyor line and the bottle unscrambling station. The tissue culture bottles on the loading conveyor line are transferred from the tissue culture bottle frame to the bottle unscrambling station in batches through the bottle loading robot arm;

[0010] The bottle unscrambling station is equipped with bottle unscrambling equipment, which is used to temporarily accommodate tissue culture bottles and sequentially input the tissue culture bottles to the bottle pouring station;

[0011] The bottle inverting station is used to detect the tissue culture bottles passing by through the image recognition algorithm, mark the tissue culture bottles with overturned posture, and push the overturned tissue culture bottles out through the execution component after the overturned tissue culture bottles move to the preset position;

[0012] The degreasing station is equipped with a degreasing device, which is arranged at the output end of the bottle inverting station and is used to receive the tested tissue culture bottles and perform degreasing operations on them;

[0013] The external washing station is provided with an external washing device, which is arranged at the output end of the degreasing device and is used to receive the degreasing tissue culture bottle and perform external washing operations on it;

[0014] The internal washing station is provided with an internal washing device, which is arranged at the output end of the external washing device and is used to receive the tissue culture bottles after external washing and perform internal washing operations on them;

[0015] The filling station is provided with a filling device, which is arranged at the output end of the internal washing device and is used to receive the tissue culture bottles after internal washing and perform filling operations on them;

[0016] The capping station is provided with a capping device, which is arranged at the output end of the filling device and is used to receive the filled tissue culture bottles and perform capping and capping operations on them;

[0017] The capping station is provided with a capping device, which is arranged at one input end of the capping device and is used to provide bottle caps for tissue culture bottles;

[0018] The upper frame conveyor line is connected to the feeding conveyor line and is used to transport the material frame of the tissue culture bottle taken away from the feeding conveyor line to the side of the transfer station to wait for being taken away;

[0019] The unloading station is equipped with an unloading robot arm, the working range of which covers the end of the upper frame conveyor line, the transfer station and the output end of the upper cover equipment, and is used to transfer the material frame on the upper frame conveyor line to the transfer station and transfer the tissue culture bottles after capping in batches to the material warehouse of the transfer station;

[0020] The transfer station is equipped with an AGV trolley, which transfers the material frame filled with tissue culture bottles to the sterilization station;

[0021] The sterilization station is equipped with a sterilization pot for sterilizing tissue culture bottles.

[0022] Furthermore, a bottle pouring station is also provided between the outer washing station and the inner washing station.

[0023] Furthermore, the material frames on the AGV are stacked in one or more layers.

[0024] Furthermore, the bottle pouring station includes a conveying track, a visual inspection mechanism arranged above the conveying track, an actuator arranged on one side of the conveying track, and an output track coordinated with the actuator. An opening is provided on the conveying track, which is connected to the top input port of the output track, and the actuator is arranged on the opposite side of the opening.

[0025] Furthermore, the visual processing method carried by the visual inspection mechanism specifically includes the following steps:

[0026] Collect images;

[0027] Preprocess the collected images;

[0028] Perform feature extraction and matching on the preprocessed image;

[0029] Send the matching results to the execution component.

[0030] Furthermore, in the image acquisition step of the visual processing method, the camera parameters are optimized in real time by the speed of the conveyor track and the ambient light intensity. When the speed of the conveyor track increases, the shutter speed is reduced in the same proportion, and when the ambient light intensity decreases, the ISO is increased and the aperture is correspondingly expanded.

[0031] In the preprocessing step, filtering is performed through an adaptive composite filtering framework, and then an adaptive gamma correction Retinex algorithm is used to make the bottle contour texture distinct and the background separation clear, so as to improve the image visual quality and processing efficiency.

[0032] Furthermore, in the feature extraction and matching steps of the visual processing method, the extracted features include shape features, texture features and edge features. The shape features quantify the regularity of the bottle by measuring the perimeter, area and circularity. The texture features use the gray-level co-occurrence matrix to calculate the contrast, correlation, energy and entropy to evaluate the complexity of the surface texture. The edge features use the Canny operator to extract the edge contour, and the Hough transform is used to fit the circle or line to detect the posture of the bottle. Then, the shape features, texture features and edge features are combined to construct a vector representing the posture of the bottle.

[0033] Matching is done by inputting the vector representing the bottle's posture into the trained model, and using the model to output the matching result;

[0034] Among them, the trained model was trained by annotating upright and overturned tissue culture bottle image samples.

[0035] Furthermore, in the feature extraction and matching steps of the visual processing method, the pixel point set of the tissue culture bottle contour is identified by an image contour detection algorithm, the contour is tracked by a chain code method, the position relationship of adjacent pixels is encoded, and the circumference value is obtained by accumulating the encoding length;

[0036] Based on the contour pixels, the area is calculated using Green's formula or integration method;

[0037] The circularity is calculated based on the circumference and area according to the circularity formula.

[0038] Furthermore, in the feature extraction and matching steps of the visual processing method, the trained model is constructed and trained by a support vector machine or a convolutional neural network.

[0039] In a second aspect, the present invention provides a control method for a fully automatic tissue culture bottle filling line, comprising the following steps:

[0040] S00, the tissue culture bottle frame to be cleaned is input into the working range of the bottle loading station through the feeding conveyor line, and the tissue culture bottles are transferred in batches to the bottle unscrambling equipment of the bottle unscrambling station through the bottle loading robot arm of the bottle loading station; when the tissue culture bottles are completely taken away from the material frame of the tissue culture bottle frame, they are transported to the side of the transfer station through the upper frame conveyor line;

[0041] S10, the tissue culture bottles are sequentially input to the bottle inverting station by the bottle unscrambling equipment; the material frame on the upper frame conveyor line is transferred to the transfer station by the unloading robot arm of the unloading station;

[0042] S20, detecting the posture of the tissue culture bottles passing by through the bottle inverting station, marking the tissue culture bottles with an overturned posture, and pushing the overturned tissue culture bottles out through the execution component after the overturned tissue culture bottles move to a preset position;

[0043] S30, the tissue culture bottles that have passed the test are degreased by the degreasing equipment at the degreasing station;

[0044] S40, the tissue culture bottle after the degreasing operation is subjected to an external cleaning operation by an external washing device at an external washing station;

[0045] S50, the tissue culture bottle after the external cleaning operation is subjected to an internal cleaning operation by an internal cleaning device of an internal cleaning station;

[0046] S60, the tissue culture bottle after the internal cleaning operation is filled by the filling equipment of the filling station;

[0047] S70, the tissue culture bottle after the filling operation is capped and screwed by the capping equipment of the capping station, and the capping equipment provides the bottle cap through the capping equipment of the capping station;

[0048] S80, the capped tissue culture bottles are transferred in batches to the material frame of the transfer station through the unloading robot arm of the unloading station;

[0049] S90, the material frame filled with tissue culture bottles is transferred to the sterilization station by the AGV cart at the transfer station to wait for sterilization operation.

[0050] The main contributions and innovations of the present invention are as follows:

[0051] 1. Efficient and automated processing: The present invention constructs a complete fully automatic tissue culture bottle filling line, integrating the whole process stations of loading, cleaning, filling, capping, unloading, transportation and sterilization. Each station works closely together and is automatically connected. There is no need for frequent manual intervention from tissue culture bottle loading to sterilization. Compared with the existing technology, it greatly improves the processing efficiency, shortens the tissue culture bottle processing cycle, strongly supports the needs of large-scale factory production of seedlings, stabilizes product quality consistency, and reduces labor costs and human error rates.

[0052] 2. Accurate bottle-falling processing mechanism: Innovatively set up the bottle-falling station, using advanced image recognition algorithms and efficient execution component combinations. Real-time monitoring of the tissue culture bottle posture, accurate identification of overturned bottles and automatic removal, avoiding interference with the production process, ensuring smooth and stable operation of the assembly line. This function is a missing link in existing technologies, effectively enhancing the degree of production automation and reliability, reducing equipment failures and product losses caused by overturned bottles, and improving overall production efficiency and yield rate.

[0053] 3. Optimize the visual inspection process: The visual inspection mechanism of the bottle inverted station intelligently optimizes the camera parameters according to the conveying speed and ambient light intensity during image acquisition to ensure clear images; the pre-processing uses adaptive composite filtering and gamma correction Retinex algorithm to deeply improve the image quality, accurately extract shape, texture, and edge features to construct the bottle posture vector, and accurately identify the bottle posture with the help of accurate training models. Compared with existing technologies, the detection accuracy and adaptability are significantly improved, laying a solid foundation for the accurate processing of inverted bottles, improving the intelligent level of the assembly line and production stability, and performing well in complex production environments.

[0054] The details of one or more embodiments of the invention are set forth in the following drawings and description to make other features, objects, and advantages of the invention more readily apparent. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0056] Figure 1 is a top view of a fully automatic tissue culture bottle filling line according to an embodiment of the present invention;

[0057] Figure 2 It is a structural schematic diagram of another implementation manner of a fully automatic tissue culture bottle filling production line according to an embodiment of the present invention;

[0058] Figure 3 is a three-dimensional diagram of a bottle inverting station according to an embodiment of the present invention;

[0059] Figure 4 yes Figure 3 A is an enlarged view of the middle image.

[0060] In the figure, 1. Operator; 2. Loading conveyor line; 3. Frame conveyor line; 4. Bottle sorting station; 5. Bottle loading station; 6. Bottle inverting station; 7. Degreasing station; 8. External washing station; 9. Internal washing station; 10. Filling station; 11. Capping station; 12. Capping station; 13. Unloading station; 14. Transfer station; 15. Sterilization station; 16. Bottle loading robot arm; 17. Unloading robot arm; 18. AGV trolley; 61. Conveying track; 62. Visual inspection mechanism; 63. Actuator; 64. Output track; 65. Guide rod; 66. Blocking limit cylinder. DETAILED DESCRIPTION

[0061] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of this specification. Instead, they are merely examples of devices and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0062] It should be noted that: in other embodiments, the steps of the corresponding method are not necessarily performed in the order shown and described in this specification. In some other embodiments, the steps included in the method may be more or less than those described in this specification. In addition, a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may be combined into a single step for description in other embodiments.

[0063] Embodiment 1

[0064] The present invention is directed to a fully automatic tissue culture bottle filling line, specifically, referring to Figure 1 ,include:

[0065] A feeding conveyor line 2 is used to input tissue culture bottle frames to be cleaned, each tissue culture bottle frame includes a material frame and a tissue culture bottle placed in the material frame;

[0066] The bottle loading station 5 is arranged at one side of the end of the feeding conveyor line 2. The bottle loading station 5 is provided with a bottle loading robot arm 16 whose working range covers the end of the feeding conveyor line 2 and the bottle unscrambling station 4. The tissue culture bottles on the feeding conveyor line 2 are transferred from the tissue culture bottle frame to the bottle unscrambling station 4 in batches through the bottle loading robot arm 16;

[0067] Bottle unscrambling station 4 is equipped with bottle unscrambling equipment for temporarily accommodating tissue culture bottles and sequentially inputting the tissue culture bottles into bottle inverting station 6;

[0068] The bottle inverting station 6 is used to detect the tissue culture bottles passing by through an image recognition algorithm, and mark the tissue culture bottles with an overturned posture, and push the overturned tissue culture bottles out through the execution component 63 after the overturned tissue culture bottles move to a preset position;

[0069] The degreasing station 7 is provided with a degreasing device, which is arranged at the output end of the bottle inverting station 6 and is used to receive the tissue culture bottles after inspection and perform degreasing operations on them;

[0070] The external washing station 8 is provided with an external washing device, which is arranged at the output end of the degreasing device and is used to receive the degreased tissue culture bottle and perform external washing operations on it;

[0071] The internal washing station 9 is provided with an internal washing device, which is arranged at the output end of the external washing device, and is used to receive the tissue culture bottle after external washing and perform internal washing operation on it;

[0072] The filling station 10 is provided with a filling device, which is arranged at the output end of the internal washing device and is used to receive the tissue culture bottles after internal washing and perform filling operations on them;

[0073] The capping station 12 is provided with a capping device, which is arranged at the output end of the filling device and is used to receive the filled tissue culture bottles and perform capping and capping operations on them;

[0074] The capping station 11 is provided with a capping device, which is arranged at one input end of the capping device and is used to provide bottle caps for tissue culture bottles;

[0075] The upper frame conveyor line 3 is connected to the upper material conveyor line 2 and is used to transport the material frame of the tissue culture bottle taken away from the upper material conveyor line 2 to the side of the transfer station 14 to wait for being taken away;

[0076] The unloading station 13 is provided with an unloading robot arm 17, the working range of which covers the end of the upper frame conveyor line 3, the transfer station 14 and the output end of the upper cover device, and is used to transfer the material frame on the upper frame conveyor line 3 to the transfer station 14 and transfer the tissue culture bottles after capping in batches to the material frame of the transfer station 14;

[0077] The transfer station 14 is provided with an AGV trolley 18, through which the material frames filled with tissue culture bottles are transferred to the sterilization station 15; preferably, the material frames on the AGV trolley 18 are stacked in one or more layers to fully utilize the AGV trolley 18.

[0078] The sterilization station 15 is provided with a sterilization pot for sterilizing the tissue culture bottles.

[0079] Among them, the improvement of the present invention is to add a feeding conveyor line 2, an upper frame conveyor line 3, a bottle loading robot arm 16, a material unloading robot arm 17, an AGV trolley 18 and a bottle inverting station 6. The equipment of the remaining stations are all prior art, such as CN117502244A or unpublished but applied patents, or other equipment disclosed on the market. Therefore, the structure and principle of the bottle unscrambling equipment, degreasing equipment, external washing equipment, internal washing equipment, filling equipment, capping equipment, capping equipment and autoclave are not described in detail. Similarly, although the feeding conveyor line 2, the upper frame conveyor line 3, the bottle loading robot arm 16, the material unloading robot arm 17, and the AGV trolley 18 are also newly added structures, they are also existing products that can be purchased, so the structure and principle are not described in detail.

[0080] like Figure 2 As shown, if canning operation is not required, the filling station 10 can be directly removed, and the capping and screwing operations can be directly performed after the internal washing operation is completed. Figure 1 and Figure 2 The equipment of the bottle sorting station 4 is different, but the basic principle is the same. The function is to input the tissue culture bottles into the assembly line in sequence. The principle and structure will not be described here, which belongs to the application of existing equipment.

[0081] In this embodiment, if Figure 3 and Figure 4 As shown, the bottle inverting station 6 includes a conveying track 61, a visual inspection mechanism 62 arranged above the conveying track 61, an actuator 63 arranged on one side of the conveying track 61, and an output track 64 coordinated with the actuator 63. The conveying track 61 is provided with an opening, which is connected to the top input port of the output track 64, and the actuator 63 is arranged on the opposite side of the opening. Among them, a guide rod 65 for guiding the tissue culture bottle is provided on the conveying track 61, so that the actuator 63 has a gap to pass through and push out the overturned tissue culture bottle. Preferably, the actuator 63 here can adopt a cylinder (bottle pushing cylinder) or an electric cylinder. If a cylinder is adopted, it is controlled by an electromagnetic valve, and the overturned tissue culture bottle can also be blocked by adding a blocking limit cylinder 66. A preferred solution is as follows:

[0082] First, input the posture photos of the tissue culture bottles in normal state and the posture photos of the overturned tissue culture bottles into the recognition system in advance; then use the industrial array camera to detect the posture state of the tissue culture bottles passing through the current conveying track 61 in real time. When the bottle is overturned, the bottle overturning signal is output to the control system, and the control system controls the blocking limit cylinder 66 to extend to block the overturned bottle; when the blocking limit cylinder 66 is extended to the right position, the control system controls the bottle pushing cylinder to push out the overturned tissue culture bottle to the output track 64; then control the bottle pushing cylinder to retract, and the subsequent bottles continue to move forward, and continue to identify whether the next tissue culture bottle is overturned through the industrial array camera. If the posture of the next tissue culture bottle is normal, the control system controls the blocking limit cylinder 66 to retract, and the subsequent tissue culture bottles continue to be smoothly transported forward. The specific input points of the control logic are shown in Table 1 below:

[0083] Table 1

[0084]

[0085] The output points are shown in Table 2:

[0086] Table 2

[0087]

[0088] Control logic:

[0089] Step 1: The host computer controls the Y3 motor to rotate forward, driving the tissue culture bottle to move forward. The industrial area array camera takes real-time photos of the position of the bottle on the conveyor track 61. At this time, the X2 point always remains in the triggered state.

[0090] Step 2: When the X1 point is triggered, the bottle falls over at the conveying track 61, and the upper computer controls the solenoid valve Y2 to be energized, and the solenoid valve moves to control the piston of the blocking limit cylinder 66 to be ejected, preventing the tissue culture bottle from continuing to move forward;

[0091] Step 3: When the X5 point is triggered, the blocking limit cylinder 66 is pushed out into place, the upper computer controls the solenoid valve Y1 to be energized, the bottle pushing cylinder piston is pushed out, and the overturned tissue culture bottle is pushed onto the output track 64. At the same time, the upper computer controls Y4 to be energized and rotate forward, driving the tissue culture bottle on the output track 64 into the collection frame in real time.

[0092] Step 4: When the X3 point is triggered, the piston of the bottle-pushing cylinder is pushed out to the position, and the tissue culture bottle has been transferred to the output track 64; then the host computer controls Y1 to lose power, the solenoid valve is reset, and the piston of the bottle-pushing cylinder is retracted.

[0093] Step 5: When the X4 point is triggered, the bottle push cylinder retracts into place. The industrial array camera continues to detect the status of the tissue culture bottle in the conveying track 61. When the X1 point is not triggered and the X2 point is triggered, the host computer controls the Y2 solenoid valve to lose power and reset the solenoid valve. At this time, the piston of the blocking limit cylinder 66 retracts; when the X6 point is triggered, the blocking limit cylinder 66 retracts into place, and the conveying track 61 can pass smoothly.

[0094] Step 6: Repeat steps 1 to 5 to detect in real time whether there are bottles on the conveying track 61, and perform precise removal actions.

[0095] In this embodiment, the visual processing method specifically includes the following steps:

[0096] Step 1: Collect images:

[0097] The industrial area array camera is the core acquisition device. Its resolution, frame rate and field of view are precisely determined according to the size of the tissue culture bottle and the speed of the conveyor track 61. For example, when the diameter of the tissue culture bottle is 80mm and the speed of the conveyor track 61 is 0.5m / s, a camera with a resolution of 2048×1536 pixels, a frame rate of 30fps, and a field of view covering a width of 300mm on the conveyor track 61 is selected to ensure clear capture of the bottle details and movement trajectory without smear and blur. It is installed 400mm above the conveyor track 61, with the optical axis perpendicular to the conveyor track 61, and the captured image is transmitted to the image processing unit.

[0098] Preferably, the camera parameters can also be dynamically adjusted according to the speed of the conveying track 61 and the ambient light intensity:

[0099] Light intensity and speed sensors are installed at key nodes of the production line to monitor parameters in real time (accuracy ±2%) and transmit them to the camera control system. According to the speed change (±10%), the shutter speed is dynamically adjusted by the PID controller. When the speed increases by 10%, the shutter is shortened from 1 / 500 second to 1 / 1100 second, ensuring that the bottle movement is frozen and the image is clear and sharp; when the light intensity fluctuates (±300lux), the fuzzy logic controller coordinates the ISO and aperture. For example, when the light intensity drops by 300lux, the ISO increases from 800 to 2000 and the f / 2.8 aperture is expanded to f / 4.5, ensuring that the image is properly exposed and the details are complete, accumulating momentum for subsequent precise detection.

[0100] Step 2: Preprocess the collected images:

[0101] The collected images are grayed, filtered to reduce noise, and enhanced in the image processing unit. The graying is done by weighted averaging, and weights are assigned according to the sensitivity of the human eye to the RGB components (such as R: 0.299, G: 0.587, B: 0.114) to reduce the amount of data. The median filter removes the salt and pepper noise, and the 3×3 template scans the image. The pixel values ​​are sorted and the median is taken to replace the central pixel. The histogram equalization enhances the contrast and stretches the gray value distribution to highlight the difference between the bottle outline and the background, laying the foundation for subsequent feature extraction.

[0102] Preferably, an adaptive composite filtering framework can also be used: after the median filter initially removes the salt and pepper noise, the wavelet threshold denoising is layered according to the noise level, the high-frequency sub-band soft threshold shrinkage noise reduction, and the low-frequency sub-band edge-preserving enhancement; for weak noise, bilateral filtering is weighted according to the pixel grayscale and spatial distance to ensure that the edge texture is finely removed while the noise is removed, and multiple algorithms work together to provide a stable image environment for feature extraction.

[0103] Preferably, the Retinex enhancement algorithm can also be advanced: introducing the adaptive gamma correction Retinex (AGCWR) algorithm, adaptively adjusting parameters according to the local and global contrast of the image, weighted calculation of the illumination component by the center surround function, dynamically stretching the reflection component to enhance detail contrast, suppress halo artifacts, make the bottle contour texture distinct, and the background separation clear, thereby improving the image visual quality and processing efficiency.

[0104] Step 3: Extract and match features of the preprocessed image:

[0105] After processing, the image is used to extract features such as shape, texture, and edge. Shape features include perimeter, area, and circularity (4π×area / perimeter). 2 ) quantify the regularity of the bottle; the texture features use the gray level co-occurrence matrix to calculate contrast, correlation, energy, and entropy to evaluate the complexity of the surface texture; the edge features are detected by the Canny operator, and the Hough transform is used to fit the straight line to locate the straight line segment parameters of the bottle contour to determine the posture. Multiple features are combined to construct a vector that characterizes the bottle's posture, such as [circularity, contrast, edge straight line slope 1, edge straight line slope 2,...].

[0106] Model training is implemented: a large number of labeled upright and overturned tissue culture bottle images are divided into training and test sets (7:3 ratio). Support vector machine (SVM), convolutional neural network (CNN) and other classification models are selected. SVM selects radial basis function kernels according to feature vectors, and adjusts penalty parameters and kernel coefficients to achieve high classification accuracy; CNN constructs a convolution, pooling, and fully connected layer architecture, such as 3 convolution layers (convolution kernels 3×3, 5×5, 3×3), 2 pooling layers (2×2 windows) and 2 fully connected layers, and uses back propagation algorithm for training and optimization. After verification by the test set, the optimal model is selected and stored in the recognition system, and the camera image vector input is received in real time to determine the bottle body posture. The processed features are input into the model, and the results can be matched.

[0107] Among them, the extraction and processing of shapes are as follows:

[0108] Calculation of perimeter C: Use image contour detection algorithm (such as findContours function of OpenCV library) to identify the pixel point set of the tissue culture bottle contour. Use chain code method or Freeman chain code to track the contour, encode according to the position relationship of adjacent pixels, and accumulate the encoding length to obtain the perimeter value. For example, if the contour pixels are encoded as 0 (horizontal right shift), 1 (upper right diagonal shift), 2 (vertical upward shift), etc., the total length of the chain code is calculated by traversing the contour, that is, the perimeter. This value helps to evaluate the regularity of the bottle shape and distinguish bottles of different specifications.

[0109] Area S determination: Based on the contour pixels, the area is calculated using the Green's formula or the integral method. The area enclosed by the contour is divided into multiple small triangles or trapezoids, and the area of ​​each sub-area is calculated and added to get the area of ​​the bottle body occupying the image pixel area. The Green's formula converts the area calculation into an integral along the contour line in a two-dimensional plane, which improves the calculation efficiency and accuracy, and provides basic data for judging the posture of the bottle body, because tipping over will cause the area projection to change.

[0110] Derivation of circularity: Based on the circumference C and area S, the formula circularity = 4πS / C 2 Calculated. A circularity close to 1 indicates that the bottle shape is close to a standard circle, and a deviation from 1 indicates an irregular shape or abnormal posture. This value sets the threshold benchmark for subsequent posture determination. If the circularity is lower than 0.9, it may indicate that the bottle is tilted or overturned, triggering further detection and confirmation.

[0111] Among them, the texture feature extraction and processing method is as follows:

[0112] Grayscale co-occurrence matrix construction: select a specific direction (such as horizontal, vertical, 45°, 135°) and displacement distance (set to 1-5 pixels according to the bottle texture scale), count the co-occurrence frequency of the pixel grayscale value in the image under this direction and distance, and construct a square matrix. Take the horizontal displacement of 1 pixel to the right as an example, traverse the image pixels, record the number of occurrences of grayscale value j at the pixel to the right of pixel i, and store it in the matrix , the matrix reflects the spatial distribution and variation characteristics of texture grayscale.

[0113] Characteristic parameter calculation: Calculate contrast, correlation, energy, entropy and other parameters from the gray level co-occurrence matrix. Contrast measures the image clarity and texture depth changes, and the formula is:

[0114] ;

[0115] Correlation represents the linear relationship between pixel grayscales, and the formula is:

[0116] ;

[0117] ( is the row mean, is the variance);

[0118] Energy reflects the uniformity and thickness of texture, and the formula is:

[0119] ;

[0120] Entropy measures texture randomness and information content, and the formula is:

[0121] ;

[0122] The various parameters work together to describe the texture characteristics of the bottle surface, which helps to distinguish between upright and overturned bottles (for example, overturning causes texture stretching and deformation, which changes the parameters), adding a multi-dimensional basis for posture judgment.

[0123] Among them, the edge feature extraction and processing method is as follows:

[0124] Canny operator edge detection: Use Gaussian filter to smooth the image and reduce noise. Set high and low thresholds according to the grayscale difference between the tissue culture bottle and the background (for example, when the background is uniformly low grayscale and the bottle body is high grayscale, the low threshold is 50 and the high threshold is 150). Use Canny operator to detect edge points and connect edge points into contours using double threshold hysteresis method. First, find strong edge pixels according to the high threshold, and connect weak edges with low threshold pixels in its neighborhood to complete the contour. Accurately locate the edge pixel set of the bottle body contour, suppress false edges with noise, and provide clear edge information for subsequent analysis.

[0125] Hough transform straight line fitting: Map the edge pixel coordinate space to the parameter space (polar coordinates ), voting and counting intersection points determine the straight line parameters. For example, the horizontal straight line corresponds to Or 180° and different ρ value point sets, if the vote exceeds the threshold, it is determined that there is a straight line, and the straight line segment of the bottle contour is fitted to obtain parameters such as slope and intercept. The posture of the bottle is analyzed by combining multiple straight line parameters. If the slopes of multiple straight lines are similar and distributed horizontally or vertically, the possibility of the bottle standing upright is high; abnormal slopes or messy distribution indicate posture changes, providing key geometric clues for posture detection to accurately determine the state of the bottle.

[0126] Step 4: Send the matching result to the execution component 63.

[0127] If the result obtained by the execution component 63 is an overturned state, the overturned tissue culture bottle is pushed out of the conveying track 61. If the result obtained is an upright state, no processing is performed.

[0128] Embodiment 2

[0129] Based on the same concept, the present invention also proposes a control method for a fully automatic tissue culture bottle filling line, comprising the following steps:

[0130] S00, the tissue culture bottle frame to be cleaned is input into the working range of the bottle loading station 5 through the feeding conveyor line 2, and the tissue culture bottles are transferred in batches to the bottle unscrambling device of the bottle unscrambling station 4 through the bottle loading robot arm 16 of the bottle loading station 5; when the tissue culture bottles are completely taken away from the material frame of the tissue culture bottle frame, they are transported to the side of the transfer station 14 through the upper frame conveyor line 3;

[0131] S10, the tissue culture bottles are sequentially input to the bottle inverting station 6 by the bottle unscrambling equipment; the material frame on the upper frame conveyor line 3 is transferred to the transfer station 14 by the unloading robot 17 of the unloading station 13;

[0132] S20, the posture of the tissue culture bottles passing by is detected by the bottle-turning station 6, and the tissue culture bottles with overturned posture are marked, and the overturned tissue culture bottles are moved to the preset position and pushed out by the execution component 63;

[0133] S30, the tissue culture bottles that have passed the test are degreased by the degreasing equipment at the degreasing station 7;

[0134] S40, the tissue culture bottle after the degreasing operation is subjected to an external cleaning operation by the external washing equipment of the external washing station 8;

[0135] S50, the tissue culture bottle after the external cleaning operation is subjected to an internal cleaning operation by the internal cleaning equipment of the internal cleaning station 9;

[0136] S60, the tissue culture bottle after the internal cleaning operation is filled by the filling equipment of the filling station 10;

[0137] S70, the tissue culture bottle after the filling operation is capped and screwed by the capping device of the capping station 12, and the capping device provides the bottle cap through the capping device of the capping station 11;

[0138] S80, the capped tissue culture bottles are transferred in batches to the material frame of the transfer station 14 through the unloading robot arm 17 of the unloading station 13;

[0139] S90, the material frame filled with tissue culture bottles is transferred to the sterilization station 15 by the AGV trolley 18 of the transfer station 14 to wait for the sterilization operation.

[0140] Preferably, the overall process of the pipeline is demonstrated as follows:

[0141] Equipment preparation and startup: In the seedling factory workshop, the equipment of each station is arranged according to the layout of the present invention. The feeding conveyor line 2, bottle loading robot arm 16, bottle sorting station 4, bottle inversion station 6 (including visual inspection and rejection mechanism), degreasing station 7, external washing station 8, internal washing station 9, filling station 10, capping station 12, capping station 11, frame conveyor line 3, unloading robot arm 17, transfer AGV trolley 18 and sterilization station 15 are ready in sequence. Start the system, and each device will enter standby after self-checking. Each conveyor line will run at a uniform speed, the robot arm will return to its position, and the visual inspection system will initialize and load the model parameters.

[0142] From feeding to bottle unscrambling: Operator 1 places 10 material frames loaded with 300 tissue culture bottles at the starting end of feeding conveyor line 2. The conveyor line moves it to bottle loading station 5 at a speed of 0.3m / s. Bottle loading robot 16 (the working range accurately covers the end of feeding conveyor line 2 and the entrance of bottle unscrambling station 4, and the stroke accuracy is ±2mm) plans the path according to the program, and grabs 300 tissue culture bottles in order and places them on the bottle unscrambling channel of bottle unscrambling station 4 within 10 minutes. The bottle unscrambling equipment (using a flexible dial mechanism, with a speed of 10r / min) guides the tissue culture bottles into the main conveyor line of bottle inversion station 6 at a constant rate, and the bottle spacing is kept at 15cm to ensure collision-free stacking.

[0143] Multi-station processing flow: The visual inspection system (industrial array camera with a resolution of 1920×1080, a frame rate of 25fps, and a field of view covering a 250mm wide area of ​​the conveyor line) at the bottle inverting station 6 monitors in real time. Normal bottles pass smoothly, and the degreasing station 7 (with 4 sets of ultrasonic transducers, a frequency of 40kHz, and a power of 200W) receives and removes grease by ultrasonic cavitation for 6 minutes. The external washing station 8 (4 sets of brushes with a speed of 120r / min) uses a brush to clean the outer wall for 5 minutes, and the internal washing station 9 (8 sets of brushes, a cylinder thrust of 50N, and a feed speed of 5mm / s) cleans the inner wall for 8 minutes. The filling station 10 (vacuum pump pumping speed of 10L / min, sensor timing accuracy of ±0.1s) completes the filling in 5 minutes according to the set amount of 100ml and the accuracy. The capping station 12 (left and right belt speed of 60r / min) cooperates with the capping station 11 (cap material channel vibration frequency of 30Hz) to complete the capping in 4 minutes.

[0144] Unloading, transfer and sterilization: After capping, the bottle is moved to the material frame of the transfer station 14 by the unloading robot arm 17 (load capacity 10kg, repeated positioning accuracy ±1mm) in 7 minutes. The transfer AGV trolley 18 (load 500kg, navigation accuracy ±10mm) transports the full frame (30 bottles per frame, stacked in 3 layers) to the sterilization station 15 (sterilization pot heating rate 5℃ / min, maintain a sterilization cycle of 20 minutes after reaching 121℃) according to the planned path in 12 minutes to wait for sterilization, completing the complete cycle of the assembly line. The whole process is closely connected, stable and efficient, providing solid automation support for tissue culture bottle processing.

[0145] Embodiment 3

[0146] This embodiment provides a solution for analyzing the bottle falling processing scenario, including:

[0147] Bottle inversion detection and response: During production, tissue culture bottles travel on the conveyor line from the internal washing to the filling room. At a certain moment, a bottle overturned due to a bottle collision. The visual inspection system instantly captures the image (acquisition exposure time 1 / 120s, gain 1.5), processes it (grayscale, 3×3 median filtering, histogram equalization takes 0.2s), matches it with the pre-stored inverted bottle posture template (roundness threshold 0.8 determines inverted bottle), and triggers the bottle inversion signal (X1 point). The control system responds within 0.5 seconds, drives the blocking limit cylinder 66 (cylinder diameter 32mm, stroke 100mm, response time 0.3s) to extend, and stops the bottle in 2 seconds. At the same time, the conveyor line motor slows down by 20% to prevent subsequent bottles from backlog.

[0148] Inverted bottle removal process: When the blocking is in place, point X5 is triggered, and the control system energizes the solenoid valve of the bottle pushing cylinder (cylinder diameter 25mm, stroke 80mm, thrust 40N). The cylinder is pushed out in 0.4 seconds, and the inverted bottle is pushed to the inverted bottle conveyor line (bandwidth 80mm, belt speed 0.3m / s) at a speed of 0.2m / s. When the bottle is pushed in place, point X3 is triggered, the cylinder solenoid valve loses power, and the cylinder is retracted back to its position in 0.3 seconds. The motor of the inverted bottle conveyor line rotates forward for 5 seconds to send the bottle into the collection frame (capacity 20). During this period, the industrial array camera continues to monitor. If the posture of the new bottle is normal, point X2 is continuously triggered. When point X6 indicates that the blocking cylinder is retracted into place (retracted in 0.4 seconds), the conveyor line motor resumes its original speed to ensure the smoothness and continuity of the production line and reduce the interference and loss of inverted bottles to production.

[0149] Embodiment 4

[0150] This embodiment provides a solution for optimizing the practice of visual inspection, including:

[0151] Dynamic parameter adjustment mechanism: In the afternoon, the light intensity in the workshop gradually weakens (from 1000 lux to 600 lux) and the conveyor line speed is fine-tuned (from 0.3 m / s to 0.4 m / s). The visual inspection system responds intelligently and the camera parameters are optimized in real time. When the light intensity becomes low, the ISO increases linearly from 100 to 200, the aperture expands from F5.6 to F4, and the shutter speed is adjusted from 1 / 120s to 1 / 150s; when the line speed increases, the camera frame rate is adaptively increased to 30fps, the exposure time is shortened to 1 / 180s, and the gain is adjusted to 1.8 to ensure stable image quality (within ±5% of the clarity deviation and within ±10% of the brightness mean fluctuation), providing a reliable image source for accurate detection and improving the system's environmental adaptability.

[0152] Feature extraction and model discrimination efficiency: Taking the images of a batch of tissue culture bottles after washing as an example, the visual processing process was optimized. The shape feature extraction obtained the bottle circumference average of 250mm, area of ​​5000mm², and circularity of 0.95; the texture feature (contrast 120, correlation 0.6, energy 0.3, entropy 1.8) accurately reflects the texture characteristics of the inner wall scratches; the edge feature is extracted by the Canny operator (high and low thresholds 80, 150), and the slope deviation of the Hough transform fitting line is within ±5°. The multi-feature vector input convolutional neural network model (5000 sets of training samples, accuracy 98%), the upright bottle is accurately judged (confidence 99%) and directly circulated, and the occasional overturned bottle (circularity 0.7, texture entropy 2.2 abnormal) is accurately identified and processed, and high-precision visual inspection is used to ensure that the assembly line efficiently produces high-quality tissue culture bottle products.

[0153] Embodiment 5

[0154] This embodiment provides a preferred solution for feature extraction and matching, which is as follows:

[0155] 1) Expansion of fine shape features

[0156] Improve the coordination between Fourier descriptor and ellipse fitting: Sample the discrete points of the tissue culture bottle contour, improve the Fourier descriptor algorithm, select the first 30 coefficients according to the weighted contour curvature, and depict the details (neck curvature, bottle body curve change) and the overall ellipse trend (axis length, eccentricity). The ellipse fitting uses the least squares method, and the parameters help locate the direction and posture of the bottle body. The two are combined to accurately construct the shape feature model, such as distinguishing normal bottles from slightly deformed bottles, with an error rate of <1%.

[0157] Shape context feature embedding: traverse the contour points, draw a fan-shaped neighborhood with the point as the center, count the distribution features of the points in the neighborhood into a histogram (8 directions, 5 distance intervals), and normalize to obtain the shape context feature vector. It is good at capturing local structural relationships and supplementing the global description limitations of Fourier descriptors, such as distinguishing the types of concave or convex positions on the bottle body, enhancing the shape feature discrimination dimension, and enhancing classification reliability.

[0158] 2) Deep texture feature mining

[0159] The fusion of 2D Gabor and 3DLBP creates a texture tensor: 2D Gabor filter (scale 4-8, direction 0°-180° interval 30°) responds to the multi-scale and multi-directional texture of the bottle body, and 3DLBP analyzes the depth change of the texture from a 3D perspective (neighborhood 3x3x3, sampling radius 1.5). The response values ​​of the two are fused into a texture feature tensor, which characterizes the thickness, direction and depth change of the bottle wall texture in three dimensions, such as accurately identifying the texture abnormalities caused by scratches and stains left inside the bottle, and improving the defect detection accuracy by 20%.

[0160] Wavelet transform to analyze texture hierarchy: Discrete wavelet transform (Daubechies4 wavelet basis) decomposes the frequency domain of texture image to obtain low-frequency approximation and high-frequency detail coefficients. Hierarchical feature vectors are constructed based on energy distribution weighted coefficients to depict the multi-resolution characteristics of textures, from macroscopic contours to microscopic defects, to fully deconstruct the surface characteristics of the bottle, providing rich texture basis for classification.

[0161] 3) Enhanced deep learning feature extraction

[0162] DCNN architecture innovation and optimization: group convolution is processed by channel grouping, which reduces the amount of calculation and parameter redundancy and improves the efficiency of feature extraction; mixed pooling (combination of maximum pooling and average pooling) ensures that the feature map is rich in information. For example, after the improvement of ResNet-50 architecture, the model training speed is increased by 40%, the feature abstraction is strong, and the extraction of key semantic features is more accurate.

[0163] Multi-scale feature fusion and training enhancement: The feature pyramid network integrates low-level details and high-level semantics, enhances context perception, and helps the model locate and identify complex scene features of the bottle. Transfer learning uses ImageNet pre-trained weights for initialization to accelerate convergence and improve generalization; adversarial training builds generator-discriminator adversarial, creates diverse samples to increase model robustness, resists interference such as light and bottle wear, ensures feature discrimination stability, and reduces the misjudgment rate by 30%.

[0164] 4) Robust and accurate multi-modal feature matching decision process

[0165] The two-layer cascade classification architecture operates efficiently: it integrates shape, texture, and deep learning features into a high-dimensional multimodal vector into the first-layer Adaboost classifier. It integrates strong classifiers by training weak classifiers based on feature importance, screens key features for dimensionality reduction, reduces computational space, and increases speed by 30%. The second-layer SVM uses radial basis functions and grid search optimization parameters (C: 0.1-10, γ: 0.01-1) for accurate classification. Multiple classifiers work together to optimize classification boundaries, reduce misjudgments, and increase recall. The overall detection accuracy reaches more than 99.5%.

[0166] Intelligent control of dynamic confidence threshold: Bayesian decision theory calculates the confidence threshold based on the misjudgment loss function (assuming missed judgment loss of 3 and misjudgment loss of 1) and the sample posterior probability. When the misjudgment rate exceeds 3%, the threshold is reduced by 5% to increase sensitivity and capture abnormalities. When the correct judgment rate is high (>98% for 100 consecutive samples), the threshold is increased by 3% to shorten the judgment interval and balance detection accuracy and efficiency. Real-time monitoring of sample distribution and model performance fine-tunes the threshold, and adaptively optimizes, improving detection efficiency by 60%, ensuring efficient and accurate production.

[0167] Those skilled in the art should understand that the technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0168] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for those of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A fully automatic tissue culture bottle filling line, characterized in that: include: A material feeding conveyor line is used to input tissue culture bottle frames to be cleaned, each tissue culture bottle frame includes a material frame and a tissue culture bottle placed in the material frame; A bottle loading station is arranged at one side of the end of the feeding conveyor line. The bottle loading station is provided with a bottle loading robot arm whose working range covers the end of the feeding conveyor line and the bottle unscrambling station. The tissue culture bottles on the feeding conveyor line are transferred in batches from the tissue culture bottle frame to the bottle unscrambling station through the bottle loading robot arm. The bottle unscrambling station is equipped with bottle unscrambling equipment, which is used to temporarily accommodate tissue culture bottles and sequentially input the tissue culture bottles to the bottle pouring station; The bottle inverting station is used to detect the tissue culture bottles passing by through the image recognition algorithm, mark the tissue culture bottles with overturned posture, and push the overturned tissue culture bottles out through the execution component after the overturned tissue culture bottles move to the preset position; The degreasing station is provided with a degreasing device, which is arranged at the output end of the bottle inverting station and is used to receive the tested tissue culture bottles and perform degreasing operations on them; The external washing station is provided with an external washing device, which is arranged at the output end of the degreasing device and is used to receive the degreased tissue culture bottle and perform external washing operations on it; The internal washing station is provided with an internal washing device, which is arranged at the output end of the external washing device and is used to receive the tissue culture bottle after external washing and perform internal washing operation on it; The filling station is provided with a filling device, which is arranged at the output end of the internal washing device and is used to receive the tissue culture bottles after internal washing and perform filling operations on them; A capping station is provided with a capping device, which is arranged at the output end of the filling device and is used to receive the filled tissue culture bottles and perform capping and capping operations on them; The capping station is provided with a capping device, which is arranged at one input end of the capping device and is used to provide bottle caps for tissue culture bottles; An upper frame conveyor line is connected to the upper material conveyor line and is used to transport the material frame of the tissue culture bottle taken away from the upper material conveyor line to one side of the transfer station to wait for being taken away; The unloading station is equipped with an unloading robot arm, the working range of which covers the end of the upper frame conveyor line, the transfer station and the output end of the upper cover equipment, and is used to transfer the material frame on the upper frame conveyor line to the transfer station and transfer the tissue culture bottles after capping in batches to the material frame of the transfer station; The transfer station is equipped with an AGV trolley, which transfers the material frame filled with tissue culture bottles to the sterilization station; The sterilization station is equipped with a sterilization pot for sterilizing tissue culture bottles.

2. A fully automatic tissue culture bottle filling line as claimed in claim 1, characterized in that: A bottle inverting station is also provided between the outer washing station and the inner washing station.

3. A fully automatic tissue culture bottle filling line as claimed in claim 1, characterized in that: The material frames on the AGV trolley are stacked in one or more layers.

4. A fully automatic tissue culture bottle filling line as claimed in claim 1, characterized in that: The bottle inverting station includes a conveying track, a visual inspection mechanism arranged above the conveying track, an actuator arranged on one side of the conveying track, and an output track coordinated with the actuator. An opening is provided on the conveying track, which is connected to the top input port of the output track, and the actuator is arranged on the opposite side of the opening.

5. A fully automatic tissue culture bottle filling line as claimed in claim 4, characterized in that: The visual processing method carried by the visual detection mechanism specifically includes the following steps: Collect images; Preprocess the collected images; Perform feature extraction and matching on the preprocessed image; Send the matching results to the execution component.

6. A fully automatic tissue culture bottle filling line as claimed in claim 5, characterized in that: In the image acquisition step of the visual processing method, the camera parameters are optimized in real time according to the speed of the conveyor track and the ambient light intensity. When the speed of the conveyor track increases, the shutter speed is reduced in the same proportion, and when the ambient light intensity decreases, the ISO is increased and the aperture is correspondingly enlarged; In the preprocessing step, filtering is performed through an adaptive composite filtering framework, and then an adaptive gamma correction Retinex algorithm is used to make the bottle contour texture distinct and the background separation clear, so as to improve the image visual quality and processing efficiency.

7. A fully automatic tissue culture bottle filling line as claimed in claim 5 or 6, characterized in that: In the feature extraction and matching steps of the visual processing method, the extracted features include shape features, texture features and edge features. The shape features quantify the regularity of the bottle by measuring the perimeter, area and circularity. The texture features use the gray-level co-occurrence matrix to calculate the contrast, correlation, energy and entropy to evaluate the surface texture complexity. The edge features use the Canny operator to extract the edge contour, and the Hough transform is used to fit the circle or line to detect the posture of the bottle. Then, the shape features, texture features and edge features are combined to construct a vector representing the posture of the bottle. Matching is done by inputting the vector representing the bottle's posture into the trained model, and using the model to output the matching result; Among them, the trained model was trained by annotating upright and overturned tissue culture bottle image samples.

8. A fully automatic tissue culture bottle filling line as claimed in claim 7, characterized in that: In the feature extraction and matching steps of the visual processing method, the pixel point set of the tissue culture bottle contour is identified by an image contour detection algorithm, the contour is tracked by a chain code method, the position relationship of adjacent pixels is encoded, and the circumference value is obtained by accumulating the encoding length; Based on the contour pixels, the area is calculated using Green's formula or integration method; The circularity is calculated based on the circumference and area according to the circularity formula.

9. A fully automatic tissue culture bottle filling line as claimed in claim 7, characterized in that: In the feature extraction and matching steps of the visual processing method, the trained model is constructed and trained by a support vector machine or a convolutional neural network.

10. A control method for a fully automatic tissue culture bottle filling line according to any one of claims 1 to 9, characterized in that: The following steps are involved: S00, the tissue culture bottle frame to be cleaned is input into the working range of the bottle loading station through the feeding conveyor line, and the tissue culture bottles are transferred in batches to the bottle unscrambling equipment of the bottle unscrambling station through the bottle loading robot arm of the bottle loading station; when the tissue culture bottles are completely taken away from the material frame of the tissue culture bottle frame, they are transported to the side of the transfer station through the upper frame conveyor line; S10, the tissue culture bottles are sequentially input to the bottle inverting station by the bottle unscrambling equipment; the material frame on the upper frame conveyor line is transferred to the transfer station by the unloading robot arm of the unloading station; S20, detecting the posture of the tissue culture bottles passing by through the bottle inverting station, marking the tissue culture bottles with an overturned posture, and pushing the overturned tissue culture bottles out through the execution component after the overturned tissue culture bottles move to a preset position; S30, the tissue culture bottles that have passed the test are degreased by the degreasing equipment at the degreasing station; S40, the tissue culture bottle after the degreasing operation is subjected to an external cleaning operation by an external washing device at an external washing station; S50, the tissue culture bottle after the external cleaning operation is subjected to an internal cleaning operation by an internal cleaning device of an internal cleaning station; S60, the tissue culture bottle after the internal cleaning operation is filled by the filling equipment of the filling station; S70, the tissue culture bottle after the filling operation is capped and screwed by the capping equipment of the capping station, and the capping equipment provides the bottle cap through the capping equipment of the capping station; S80, the capped tissue culture bottles are transferred in batches to the material frame of the transfer station through the unloading robot arm of the unloading station; S90, the material frame filled with tissue culture bottles is transferred to the sterilization station by the AGV cart at the transfer station to wait for sterilization operation.

Images