Unmanned thermal battery packaging quality inspection equipment based on machine vision

Abstract

The invention provides unmanned thermal battery packaging quality inspection equipment based on machine vision, and relates to the technical field of production line detection systems.The unmanned thermal battery packaging quality inspection equipment comprises a moving module, a quality inspection module and a base station assembly, a mounting bottom plate is arranged in the middle of the moving module, and a connecting mechanical arm and a mounting support are arranged at the end of the quality inspection module in an extending mode; mechanical joints are arranged at the connecting positions of the adjacent connecting mechanical arms, the connecting positions of the connecting mechanical arms and the mounting bracket and the connecting positions of the quality inspection module and the connecting mechanical arms, a high-resolution camera and a multi-layer convolutional neural network are adopted, tiny defects can be accurately detected, and compared with a traditional manual detection method, the detection efficiency is greatly improved. According to the automatic quality inspection device, the risks of missing detection and misjudgment are remarkably reduced, the consistency of product quality is ensured, full-automatic quality inspection is achieved through unmanned design, the labor cost and human errors are reduced, and the production efficiency is improved. In addition, an intelligent navigation system and a laser radar technology enable the equipment to move and position autonomously in a complex environment, so that the working flexibility is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production line detection systems, and in particular to unmanned thermal battery packaging quality inspection equipment based on machine vision. Background Art

[0002] According to the workpiece quality inspection method and device based on machine vision disclosed in the Chinese patent publication number "CN116883417B", it relates to the field of quality inspection technology, including selecting a standard image and multiple non-standard images of each workpiece from images of several kinds of workpieces; obtaining a real feature point set of each standard image and non-standard image; dividing the multiple non-standard images into a training set and a test set; building an initial network model with a generative adversarial network, and training and testing the initial network model with the training set and the test set; inputting the image of the workpiece to be inspected into a feature point detection model to obtain a corresponding predicted feature point set; calculating a first feature difference value between the predicted feature point set of the image of the workpiece to be inspected and the real feature point set of the standard image, and judging whether the workpiece to be inspected is qualified according to the first feature difference value.

[0003] According to the machine vision quality inspection robot disclosed in Chinese patent publication number "CN114813569A", it includes a frame base, an adjustment component, a fastening component, a quality inspection component, a support table, a support plate, a flip frame, a motor base, an inspection plate, a column, a connecting block, a visual camera body and a transparent observation cover. The support table is fixedly installed on the top of the frame base, and an operating bracket is arranged on one side of the frame base. The present invention uses the set fastening component to position and clamp the product on the inspection plate to prevent the product from shifting during the inspection process, thereby improving the accuracy of the inspection. The set cleaning motor is used to drive the cleaning gear ring and the cleaning bracket to rotate, so as to automatically clean the transparent observation cover.

[0004] According to the Chinese patent publication number "CN116095099A", a mechanical parts quality inspection system based on machine vision is disclosed. By setting up a quality inspection terminal, the quality inspection of the complete mechanical parts produced and processed in the mechanical processing factory is carried out. The preparation unit generates remote hash slot rule data, the redis cluster management unit creates a redis cluster, the encryption unit generates a rejection roulette and generates a storage strategy for each confidential part quality inspection report data based on the rejection roulette, and the quality inspection storage package corresponding to the confidential part quality inspection report data of the virtual storage module is uploaded to the corresponding hash slot of the corresponding cloud server.

[0005] The above patent documents and prior art have the following technical problems when used:

[0006] Traditional quality inspection methods usually rely on manual inspection of each piece, which is slow and inefficient, and cannot meet the needs of modern large-scale production. Manual inspection is often subjective and can be easily affected by factors such as the experience and fatigue of the inspectors, leading to missed inspections and misjudgments. Summary of the invention

[0007] Technical issues solved

[0008] In view of the shortcomings of the prior art, the present invention provides an unmanned thermal battery packaging quality inspection device based on machine vision, which solves the following problems:

[0009] Traditional quality inspection methods usually rely on manual inspection of each piece, which is slow and inefficient, and cannot meet the needs of modern large-scale production. Manual inspection is often subjective and can be easily affected by factors such as the experience and fatigue of the inspectors, leading to missed inspections and misjudgments.

[0010] Technical Solution

[0011] To achieve the above objectives, the present invention is implemented through the following technical solutions: unmanned thermal battery packaging quality inspection equipment based on machine vision, including a mobile module, a quality inspection module and a base station assembly, the middle of the mobile module is provided with a mounting base plate, the end of the quality inspection module is extended with a connecting robot arm and a mounting bracket, and the connection positions of adjacent connecting robot arms, the connection positions of the connecting robot arms and the mounting bracket, and the connection positions of the quality inspection module and the connecting robot arm are all provided with mechanical joints.

[0012] Preferably, the base station assembly comprises a base station body positioned by a wall, and a connecting cable and a telescopic connecting rod are provided between the base station body and the mobile module.

[0013] Preferably, the connecting cable is wound around the tail of the mobile module, the telescopic connecting rod is built into the inner cavity of the base station body, and the ends of the connecting cable and the telescopic connecting rod are both provided with electric plugs, which adopt a pin interface.

[0014] Preferably, the quality inspection module includes a high-resolution camera, and defect detection is performed through a multi-layer convolutional neural network. The output of the multi-layer convolutional neural network is finally classified through a logistic regression model. The specific formula is as follows:

[0015] y=σ(W T ·h+b)

[0016] y is the predicted defect probability, σ is the Sigmoid activation function, defined as σ(z) represents the activation function value at position z, e is a constant, W is the weight matrix, h is the feature vector extracted by CNN, and b is the bias term.

[0017] Preferably, an adaptive light source system is provided between the high-resolution camera and the quality inspection module, and the adaptive light source system is based on adaptive threshold segmentation in the image processing algorithm, using the formula:

[0018]

[0019] T is the adaptive threshold of the image, N is the total number of pixels, and I(i) is the gray value of the i-th pixel.

[0020] Preferably, the mobile module also includes an intelligent navigation system, which realizes environmental perception and path planning through laser radar and sensors, and performs dynamic optimization in combination with a deep reinforcement learning algorithm. The specific update strategy is:

[0021]

[0022] Q(s,a) is the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ is the next state, and Q(s′,a′) is the value of taking action a′ in state s′.

[0023] Preferably, the base station component further comprises a data processing unit for processing quality inspection data in real time, using an algorithm combining a convolutional neural network and a support vector machine, combined with an optimization algorithm, and training through the following loss function:

[0024]

[0025] L is the total loss, N is the number of samples, y i is the true label, is the predicted value, λ is the regularization parameter, and w is the model weight vector;

[0026] The data processing unit adopts an integrated learning strategy and makes decisions through the following weighted voting mechanism:

[0027]

[0028] is the final prediction, K is the number of models, and w k is the weight of the kth model, h k (x) is the prediction of the kth model for input x.

[0029] Preferably, the quality inspection module further includes an adaptive feedback unit, which updates the model weights by the following formula:

[0030]

[0031] w t+1 is the updated weight, w t is the current weight, η is the learning rate, is the gradient of the current weight.

[0032] Preferably, the working steps of the unmanned thermal battery packaging quality inspection equipment are as follows:

[0033] Sp1: System initialization, the mobile module starts and performs self-test to confirm that all hardware components are normal;

[0034] Sp2: Activate the intelligent navigation system, scan the environment, and generate a moving path;

[0035] Sp3: The mobile module moves to the quality inspection area according to the navigation path to ensure correct positioning;

[0036] Sp4: The quality inspection module starts working and starts multiple high-resolution cameras for image acquisition;

[0037] Sp5: Adaptive light source system automatically adjusts light intensity according to ambient light to ensure image quality;

[0038] Sp6: Process the acquired images through convolutional neural networks for defect detection;

[0039] Sp7: The data processing unit analyzes quality inspection data in real time and optimizes the processing results through edge computing;

[0040] Sp8: Combines defect detection results with the quality assessment module to calculate the comprehensive quality score;

[0041] Sp9: Output quality inspection report and transmit the results to the monitoring system through the wireless communication module.

[0042] Preferably, the image processing algorithm in the quality inspection module includes the following steps:

[0043] Sp1: image preprocessing, including denoising, contrast enhancement and adaptive threshold segmentation;

[0044] Sp2: Extract features and classify defects through multi-layer convolutional neural networks;

[0045] Sp3: Use the logistic regression model to make the final decision and output the defect probability.

[0046] Beneficial Effects

[0047] The present invention provides an unmanned thermal battery packaging quality inspection device based on machine vision. It has the following beneficial effects:

[0048] 1. The present invention uses high-resolution cameras and multi-layer convolutional neural networks (CNNs) to accurately detect tiny defects. Compared with traditional manual detection methods, it significantly reduces the risk of missed detection and misjudgment, ensuring the consistency of product quality. Through unmanned design, the equipment realizes fully automatic quality inspection, reduces labor costs and human errors, and improves production efficiency. In addition, the intelligent navigation system and laser radar technology enable the equipment to move and position autonomously in complex environments, improving work flexibility.

[0049] 2. The present invention uses data processing units and edge computing capabilities to enable the equipment to analyze quality inspection data in real time and provide timely feedback on defect information, which facilitates timely adjustment of the production line, reduces the generation of defective products, and improves production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 It is the overall structure diagram of the device of the present invention;

[0051] Figure 2 A side view of the device of the present invention;

[0052] Figure 3 It is a front view of the device of the present invention;

[0053] Figure 4 It is a schematic diagram of the use of the device of the present invention;

[0054] Figure 5 This is a style diagram of the base station component of the present invention, with the left side being a unidirectional base station and the right side being a multi-directional base station.

[0055] Among them: 1. Mobile module; 2. Installation base plate; 3. Energy storage module; 4. Installation bracket; 5. Mechanical joint; 6. Connecting mechanical arm; 7. Quality inspection module; 8. Base station assembly; 801. Base station body; 802. Telescopic connecting rod; 803. Connecting cable. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Specific embodiment one:

[0058] As shown in the figure, the unmanned thermal battery packaging quality inspection equipment based on machine vision includes a mobile module 1, a quality inspection module 7 and a base station component 8. A mounting base plate 2 is provided in the middle of the mobile module 1. A connecting mechanical arm 6 and a mounting bracket 4 are extended from the end of the quality inspection module 7. Mechanical joints 5 are provided at the connection positions of adjacent connecting mechanical arms 6, the connection positions of the connecting mechanical arms 6 and the mounting bracket 4, and the connection positions of the quality inspection module 7 and the connecting mechanical arms 6. The mobile module 1 includes a chassis, a drive motor, a wheeled mobile system, and an intelligent navigation sensor laser radar. The quality inspection module 7 includes: multiple high-resolution cameras, an adaptive light source system, a computing unit GPU, and a multi-layer convolutional neural network CNN processor. The base station component 8 includes: a base station body 801, a data processing unit, a wireless communication module, a connecting cable 803 and a telescopic connecting rod 802.

[0059] The mobile module 1 and the mounting base plate 2 are made of aluminum alloy or high-strength plastic to provide structural strength and stability, support other components, and provide good heat dissipation performance.

[0060] Drive motor: Brushless DC motor BLDC or stepper motor, with high efficiency and long life. Provide power to the wheels and control the speed and direction of movement.

[0061] Wheeled mobile system: four driving wheels and one universal wheel, equipped with shock-absorbing devices to adapt to uneven ground. Ensure that the equipment can be moved flexibly and adapt to various environments.

[0062] Lidar sensor: Lidar Lite or Velodyne. Perform 360-degree environment scanning, generate maps and avoid obstacles.

[0063] Infrared sensor: Passive infrared sensor PIR. Detects obstacles and heat sources in front to prevent collision.

[0064] IMU inertial measurement unit: MPU-6050 or LSM6DS3. Real-time monitoring of the acceleration and angle changes of the device, providing navigation information.

[0065] Power management module: lithium battery pack, battery management system BMS. Provide power to all components, monitor battery status and ensure safety.

[0066] The high-resolution camera of the quality inspection module 7 adopts the Sony IMX series. The resolution is at least 1080p and supports high-speed shooting. It can capture images from multiple angles and is suitable for different workpieces.

[0067] Adaptive light source system: LED light, light intensity sensor. Automatically adjust light intensity according to ambient light changes to ensure image quality.

[0068] The image processing unit includes a high-performance GPU NVIDIA Jetson series or FPGA. It executes image processing and deep learning algorithms to accelerate image analysis.

[0069] Robotic Arm 6: Six-degree-of-freedom robotic arm. Adjusts the camera position and performs precise operations such as flipping or moving the inspected object.

[0070] The mounting bracket 4 is made of steel or aluminum alloy and has stability, and fixes the quality inspection module and maintains the relative position of the camera and the light source.

[0071] Base station body 801: Steel structure, ensuring durability. As the control center of the equipment, connecting all modules.

[0072] Wireless communication module: Use Wi-Fi or LoRa module to realize data transmission and remote monitoring.

[0073] Connection cable 803: Uses shielded twisted pair cable to ensure signal stability. Connects each module and transmits power and data.

[0074] Telescopic connecting rod 802: Aluminum alloy, light and strong. Connects the mobile module and the base station to provide flexible space adjustment.

[0075] The data processing unit uses a computer or server to receive data from the quality inspection module and conduct comprehensive analysis and storage.

[0076] Cooling system: Fan or heat sink. Provides cooling for electronic components to ensure stable operation of the device under high load.

[0077] User interface: Touch screen display, button panel. Provides an interface for users to interact with the device, display status and receive instructions.

[0078] The connecting cable 803 is wound around the tail of the mobile module 1, and the telescopic connecting rod 802 is built into the inner cavity of the base station body 801. The ends of the connecting cable 803 and the telescopic connecting rod 802 are both provided with electrical plugs, and the electrical plugs use pin interfaces.

[0079] The quality inspection module 7 includes a high-resolution camera, and performs defect detection through a multi-layer convolutional neural network. The output of the multi-layer convolutional neural network is finally classified through a logistic regression model. The specific formula is as follows:

[0080] y=σW T ·h+b

[0081] y is the predicted defect probability, σ is the Sigmoid activation function, defined as σ(z) represents the activation function value at position z, e is a constant, W is the weight matrix, h is the feature vector extracted by CNN, and b is the bias term.

[0082] An adaptive light source system is provided between the high-resolution camera and the quality inspection module 7. The adaptive light source system is based on the adaptive threshold segmentation in the image processing algorithm, using the formula:

[0083]

[0084] T is the adaptive threshold of the image, N is the total number of pixels, and Ii is the gray value of the i-th pixel.

[0085] The mobile module 1 also includes an intelligent navigation system, which realizes environmental perception and path planning through lidar and sensors, and combines deep reinforcement learning algorithm for dynamic optimization. The specific update strategy is:

[0086]

[0087] Qs,a is the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ is the next state, and Qs′,a′ is the value of taking action a′ in state s′. The intelligent navigation system can update the mobile path in real time and exchange data with the quality inspection module 7 through the information sharing algorithm. Specifically, the weighted mean algorithm is used, and the calculation formula is:

[0088]

[0089] is the weighted mean, w i is the weight of the ith data point, x i is the value of the ith data point.

[0090] The base station component 8 also includes a data processing unit for real-time processing of quality inspection data, using an algorithm combining a convolutional neural network with a support vector machine SVM, combined with an optimization algorithm, and training through the following loss function:

[0091]

[0092] L is the total loss, N is the number of samples, y i is the true label, is the predicted value, λ is the regularization parameter, and w is the model weight vector;

[0093] The data processing unit adopts an integrated learning strategy and makes decisions through the following weighted voting mechanism:

[0094]

[0095] is the final prediction, K is the number of models, and w k is the weight of the kth model, h k x is the prediction of the kth model for input x.

[0096] The quality inspection module 7 also includes an adaptive feedback unit, which updates the model weights by the following formula:

[0097]

[0098] w t+1 is the updated weight, w t is the current weight, η is the learning rate, is the gradient of the current weight.

[0099] The quality inspection system includes an integrated quality assessment module, which uses a multi-level decision-making algorithm for comprehensive evaluation, which is implemented through the following formula:

[0100] Q=α 1 D1 +α 2 D 2 +α 3 D 3

[0101] Q is the comprehensive quality score. 1 , D 2 , D 3 They are different types of defect detection indicators. 1 , α 2 , α 3 is the corresponding weight coefficient, which is adaptively adjusted according to historical data.

[0102] The quality assessment module optimizes the weight coefficients through a reinforcement learning algorithm, specifically using the following update formula:

[0103] α new =α old +η·(RQ)

[0104] α new is the updated weight. old is the old weight, η is the learning rate, and R is the desired quality score.

[0105] The data processing unit has real-time analysis capabilities based on edge computing and performs data fusion through the following formula:

[0106]

[0107] F is the fused data result. N is the number of sensors.

[0108] D i is the data of the i-th sensor, W i is the corresponding weight.

[0109] It also includes a remote monitoring module that can communicate with external systems through a RESTful API, using the following data transmission protocols:

[0110] P=POST(endpoint,D)

[0111] P is the sending request, endpoint is the address of the target server, D is the data to be sent, POST is the communication protocol, including but not limited to Bee, Lora and cellular network data protocols, and the protocols all use encryption algorithms for data encryption. The entire system of the entire device has the function of self-learning and improvement, and can adjust the detection parameters based on historical data and feedback through the following formula:

[0112] P new =P old +γ(EP old )

[0113] P new is the updated parameter value, P old is the old parameter value, E is the expected value, and γ is the adjustment factor.

[0114] Mobile Module:

[0115] Install the base plate, wheeled mobile platform, drive motor, lidar sensor, infrared sensor, and IMU.

[0116] The mobile module controls wheeled movement by driving motors, and the laser radar and infrared sensors scan the environment in real time, generate maps and perform path planning. The IMU provides location information to ensure high-precision navigation.

[0117] Quality inspection module:

[0118] Multiple high-resolution cameras, adaptive lighting system, image processing unit, robotic arm, mounting bracket.

[0119] The camera collects images of thermal battery packaging from multiple angles with the assistance of the light source system, and the image data is transmitted to the image processing unit. The processing unit uses a multi-layer convolutional neural network to analyze the image and detect defects. The robotic arm is responsible for positioning and adjusting the camera angle.

[0120] Base station components:

[0121] Base station body, wireless communication module, connecting cables, telescopic connecting rod, and data processing unit.

[0122] The base station exchanges data with the mobile module and quality inspection module in real time through the wireless communication module, and connects cables and telescopic rods to ensure stable power supply and signals.

[0123] 2. Operation process

[0124] Startup and initialization: The device is powered on and the system self-checks to ensure that all sensors, drive motors, and modules are functioning properly.

[0125] Environment scanning and path planning: The mobile module is started, and the lidar and sensors scan the environment, generate a map and calculate the best movement path.

[0126] Move to the quality inspection area: The mobile module moves to the target location according to the path planning, and makes real-time adjustments during the process to avoid obstacles.

[0127] Image acquisition: The quality inspection module is activated, the high-resolution camera starts to collect thermal battery packaging images, and the adaptive light source system automatically adjusts the light intensity to adapt to environmental changes.

[0128] Image processing and defect detection: The image data is transmitted to the image processing unit, and CNN is used for defect recognition.

[0129] Result output and storage: The defect detection results are transmitted to the monitoring center through the wireless communication module and stored in the data processing unit for subsequent analysis.

[0130] Maintenance and feedback: The equipment performs self-maintenance regularly, and the feedback data can be used to optimize subsequent operations and improve detection accuracy and efficiency. Specific embodiment 2:

[0132] The working steps of unmanned thermal battery packaging quality inspection equipment based on machine vision are as follows:

[0133] Sp1: System initialization, mobile module 1 starts and performs self-test to confirm that all hardware components are normal;

[0134] Sp2: Activate the intelligent navigation system, scan the environment, and generate a moving path;

[0135] Sp3: Mobile module 1 moves to the quality inspection area according to the navigation path to ensure correct positioning;

[0136] Sp4: The quality inspection module 7 starts working and starts multiple high-resolution cameras for image acquisition;

[0137] Sp5: Adaptive light source system automatically adjusts light intensity according to ambient light to ensure image quality;

[0138] Sp6: Process the collected images through convolutional neural network (CNN) to perform defect detection;

[0139] Sp7: The data processing unit analyzes quality inspection data in real time and optimizes the processing results through edge computing;

[0140] Sp8: Combines defect detection results with the quality assessment module to calculate the comprehensive quality score;

[0141] Sp9: Output quality inspection report and transmit the results to the monitoring system through the wireless communication module.

[0142] The image processing algorithm in the quality inspection module 7 includes the following steps:

[0143] Sp1: image preprocessing, including denoising, contrast enhancement and adaptive threshold segmentation;

[0144] Sp2: Extract features and classify defects through multi-layer convolutional neural network (CNN);

[0145] Sp3: Use the logistic regression model to make the final decision and output the defect probability.

[0146] During the system initialization phase, the mobile module starts a self-test program to check all hardware components, including drive motors, sensors, and communication modules, to ensure that they are working properly. Once completed, the intelligent navigation system is activated, and the lidar and infrared sensors begin to scan the environment, generate real-time maps, and calculate the best movement path based on the location of obstacles. This process ensures that the mobile module can reach the predetermined quality inspection area accurately and efficiently. During the movement, the IMU provides position information to correct any deviations.

[0147] Once the mobile module is positioned in the quality inspection area, the quality inspection module is started, and multiple high-resolution cameras are activated for image acquisition to ensure that complete images are obtained from different angles. At this time, the adaptive light source system automatically adjusts the light intensity according to the changes in ambient light, optimizes the shooting conditions, and ensures image quality.

[0148] Next, the captured images enter the processing stage, using a multi-layer convolutional neural network for feature extraction and defect detection. CNN analyzes the image based on the trained model and identifies potential defect areas. The processed data is transmitted to the data processing unit for real-time analysis, and edge computing technology is used to optimize the processing results and improve processing speed and efficiency.

[0149] Subsequently, the defect detection results are combined with the quality assessment module to calculate the comprehensive quality score. The specific scoring formula may include the weighted average of multiple indicator weights, such as dynamically adjusting the weight coefficient according to the defect type and severity. Finally, the quality inspection report is generated and output, and the results are transmitted to the monitoring system in real time through the wireless communication module to facilitate subsequent data analysis and decision-making.

[0150] The image processing algorithm in the quality inspection module includes several key steps: in image preprocessing, denoising is first performed to eliminate interference, then image enhancement technology is used to improve contrast, and finally an adaptive threshold segmentation method is used to separate the target from the background for subsequent analysis. In the feature extraction stage, a multi-layer convolutional neural network performs deep feature learning on the preprocessed image to effectively classify the defect type. Finally, in the defect judgment stage, a logistic regression model is applied to calculate the defect probability based on the extracted feature vector and output the final detection result. This series of steps constitutes an efficient quality inspection process for the entire equipment, ensuring that the quality of thermal battery packaging meets the standards. Specific embodiment three:

[0152] The hardware working logic of the unmanned thermal battery packaging quality inspection equipment based on machine vision is as follows:

[0153] This device is a freely movable quality inspection device. The quality inspection module 7 at the end thereof includes a camera with built-in machine vision, which can perform quality inspection on the products in the camera shooting area in real time. There are two types of quality inspection databases. One is the traditional design parameters, which have the drawings of standard qualified products as the comparison source data of the camera. The other part will be learned in real time through the various algorithms proposed in the specific embodiment 1 during the continuous working process, and a new online database (this database can be a cloud database or a local database) will be constructed, so that real-time visual detection can be achieved, thereby achieving good automatic error correction capabilities. It should be clear that in terms of the matching relationship between the mounting base plate 2 and the mounting bracket 4 of the whole machine, on the one hand, bolts can be directly used for fastening connection, which has low cost and low modification difficulty. Another form uses a strong magnetic mounting base plate 2, which can directly adsorb the mounting bracket 4 tightly on the surface of the mounting base plate 2, and the mounting base plate 2 is directly mounted on the mobile module 1 by bolts. The mobile module 1 is essentially a crawler chassis, which can also be changed to a wheeled chassis according to actual conditions. The conventional crawler chassis has relatively high passability and safety redundancy in the industrial environment.

[0154] At the same time, the whole machine has two operating modes:

[0155] Method 1: Continuous detection at a fixed position. The whole machine is moved to the corresponding position of the assembly line through the mobile module 1. The connecting cable 803 at the rear of the equipment is inserted into the plug of the corresponding station for uninterrupted power supply. The whole machine is fixed at the station position of the assembly line and connected to the mechanical arm 6 to drive the quality inspection module 7 to realize multi-degree-of-freedom thermal battery product quality inspection.

[0156] Method 2: Random quality inspection of the thermal battery assembly line. The whole machine is at the position of the base station component 8, and the energy storage module 3 is charged by connecting the cable 803 or the telescopic connecting rod 802. The connecting cable 803 and the telescopic connecting rod 802 are both plugs, which are directly plugged into the plug quality inspection in the inner cavity of the base station component 8 to achieve electrical connectivity. On the one hand, charging can be achieved, and on the other hand, information transmission and data backup can be achieved. Directly, the spatial expansion capability of the base station component 8 will be stronger, and the amount of data it can carry or the expanded data space will be more. The data can be stored inside the base station component 8, so that the base station component 8 constitutes a small data server. When multiple groups of equipment are required in intelligent factories and larger factory areas, the base station component 8 can also be expanded into a multi-directional central base station, see Figure 5 .

[0157] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprising a reference structure" do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0158] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Unmanned thermal battery packaging quality inspection equipment based on machine vision, characterized by: The invention comprises a mobile module (1), a quality inspection module (7) and a base station assembly (8); a mounting base plate (2) is provided in the middle of the mobile module (1); a connecting mechanical arm (6) and a mounting bracket (4) are extended from the end of the quality inspection module (7); and mechanical joints (5) are provided at the connection positions of adjacent connecting mechanical arms (6), the connection positions of the connecting mechanical arms (6) and the mounting bracket (4), and the connection positions of the quality inspection module (7) and the connecting mechanical arms (6).

2. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 1 is characterized in that: The base station assembly (8) comprises a base station body (801) positioned by a wall, and a connecting cable (803) and a telescopic connecting rod (802) are provided between the base station body (801) and the mobile module (1).

3. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 2 is characterized in that: The connecting cable (803) is wound around the tail of the mobile module (1), the telescopic connecting rod (802) is built into the inner cavity of the base station body (801), and the ends of the connecting cable (803) and the telescopic connecting rod (802) are both provided with electrical plugs, which adopt a pin interface.

4. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 1 is characterized in that: The quality inspection module (7) includes a high-resolution camera, and performs defect detection through a multi-layer convolutional neural network. The output of the multi-layer convolutional neural network is finally classified through a logistic regression model. The specific formula is as follows: y=σ(W T ·h+b) y is the predicted defect probability, σ is the Sigmoid activation function, defined as Represents the activation function value at the z position, e is a constant, W is the weight matrix, h is the feature vector extracted by CNN, and b is the bias term.

5. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 4 is characterized in that: An adaptive light source system is provided between the high-resolution camera and the quality inspection module (7). The adaptive light source system is based on adaptive threshold segmentation in the image processing algorithm and uses the formula: T is the adaptive threshold of the image, N is the total number of pixels, and I(i) is the gray value of the i-th pixel.

6. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 1 is characterized in that: The mobile module (1) also includes an intelligent navigation system, which realizes environmental perception and path planning through laser radar and sensors, and performs dynamic optimization in combination with a deep reinforcement learning algorithm. The specific update strategy is: Q(s,a) is the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, s′ is the next state, and Q(s′,a′) is the value of taking action a′ in state s′.

7. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 1 is characterized in that: The base station component (8) also includes a data processing unit for real-time processing of quality inspection data, using an algorithm combining a convolutional neural network and a support vector machine (SVM), combined with an optimization algorithm, and training through the following loss function: L is the total loss, N is the number of samples, y i is the true label, is the predicted value, λ is the regularization parameter, and w is the model weight vector; The data processing unit adopts an integrated learning strategy and makes decisions through the following weighted voting mechanism: is the final prediction, K is the number of models, and w k is the weight of the kth model, h k (x) is the prediction of the kth model for input x.

8. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 1 is characterized in that: The quality inspection module (7) also includes an adaptive feedback unit, which updates the model weights using the following formula: w t+1 is the updated weight, w t is the current weight, η is the learning rate, is the gradient of the current weight.

9. The unmanned thermal battery packaging quality inspection equipment based on machine vision according to any one of claims 1 to 8, characterized in that: The working steps of the unmanned thermal battery packaging quality inspection equipment are as follows: Sp1: System initialization, the mobile module (1) starts and performs self-test to confirm that all hardware components are normal; Sp2: Activate the intelligent navigation system, scan the environment, and generate a moving path; Sp3: The mobile module (1) moves to the quality inspection area according to the navigation path to ensure correct positioning; Sp4: The quality inspection module (7) starts working and starts multiple high-resolution cameras to collect images; Sp5: Adaptive light source system automatically adjusts light intensity according to ambient light to ensure image quality; Sp6: Process the acquired images through convolutional neural network (CNN) for defect detection; Sp7: The data processing unit analyzes quality inspection data in real time and optimizes the processing results through edge computing; Sp8: Combines defect detection results with the quality assessment module to calculate the comprehensive quality score; Sp9: Output quality inspection report and transmit the results to the monitoring system through the wireless communication module.

10. The working steps of the unmanned thermal battery packaging quality inspection equipment based on machine vision according to claim 9 are characterized in that: The image processing algorithm in the quality inspection module (7) comprises the following steps: Sp1: image preprocessing, including denoising, contrast enhancement and adaptive threshold segmentation; Sp2: Extract features and classify defects through multi-layer convolutional neural network (CNN); Sp3: Use the logistic regression model to make the final decision and output the defect probability.

Images