Capacitance proximity sensing substance identification system and method based on machine learning

Abstract

The invention provides a capacitive proximity sensing substance identification system and method based on machine learning, and belongs to the technical field of non-contact substance identification, the capacitive proximity sensing substance identification system comprises a system shell and an internal structure, the system shell is composed of a display screen and a sensing area, the internal structure is composed of a power supply module, a data acquisition module, a miniature single-board computer and a display module, the power supply module is electrically connected with the miniature single-board computer and the display module, the data acquisition module is electrically connected with the miniature single-board computer, the miniature single-board computer is electrically connected with the display module, and the display module is electrically connected with the data acquisition module. The data acquisition module is composed of a capacitance proximity sensor and a distance sensor. According to the capacitive proximity sensing material identification system and method based on machine learning, the problems of sensor abrasion, limited material identification, insufficient measurement precision and high hardware resource demand in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of non-contact material recognition, and in particular to a capacitance proximity sensing material recognition system and method based on machine learning. Background Art

[0002] Existing material recognition technologies are mainly divided into two categories: One is a material recognition system based on visual perception, and the core lies in the optimization degree of image quality and recognition algorithms. Such a system relies on visual sensors such as cameras to capture images of objects, and the quality of the images covers multiple dimensions such as resolution, contrast, and lighting conditions. The performance of the recognition algorithm is deeply affected by its complexity and accuracy. In order to further improve the recognition accuracy, the algorithm needs to be continuously optimized and improved to achieve better recognition effects.

[0003] Another mainstream material recognition method is a system based on tactile perception. In this system, the primary challenge is to ensure that the tactile sensor can quickly and accurately capture comprehensive information about the material. Such a system requires the sensor design to solve key performance indicators such as accuracy, sensitivity, and repeatability, ensuring that the sensor can work continuously and stably during high-speed and high-frequency material grasping operations. However, long-term and frequent material contact often leads to wear and deformation of the contact parts of the sensor, thereby affecting its repeatability and sensitivity.

[0004] In addition, a single tactile sensor is difficult to comprehensively capture the tactile characteristics of the material, which limits the accuracy and reliability of material recognition. To solve this problem, multiple sensors need to be deployed at different positions of the material. Although this array-type multi-sensor configuration can improve the recognition ability, it also brings high preparation costs. The data synchronization and fusion problems between multiple sensors will cause data crosstalk, further increasing the challenges of system implementation. Therefore, when designing and optimizing a material recognition system based on tactile perception, multiple aspects such as sensor performance, cost control, and data processing efficiency need to be comprehensively considered.

[0005] In the actual application of existing traditional non-contact recognition sensors, there are also significant technical limitations. Their design cannot accurately identify the distance between the material and the detected material, and at the same time, their operation highly depends on the computing power support of the computer terminal. It is necessary to train the algorithm model through a computer and then perform material recognition, and it cannot operate independently. This design limitation restricts the deployment flexibility of the device and the diversity of application scenarios, and also shows obvious deficiencies in terms of real-time performance and portability.

[0006] In summary, the substance recognition system based on visual perception highly depends on the image quality and the optimization degree of the recognition algorithm; the substance recognition system based on tactile perception faces challenges in sensor performance; the traditional non-contact recognition sensor has problems in accurately identifying the distance between the substance and the detection substance in design, highly depends on the computing power support of the computer terminal, and requires algorithm model training through a computer to perform substance recognition, and cannot operate independently. Summary of the Invention

[0007] The purpose of the present invention is to provide a capacitance proximity sensing substance recognition system and method based on machine learning, which solves the problems of sensor wear, limited substance recognition, insufficient measurement accuracy, and high hardware resource requirements in the prior art.

[0008] To achieve the above purpose, the present invention provides a capacitance proximity sensing substance recognition system based on machine learning, including a system housing and an internal structure. The system housing is composed of a display screen and a sensing area, and the internal structure is composed of a power module, a data acquisition module, a micro single-board computer, and a display module. The power module is electrically connected to the micro single-board computer and the display module respectively, the data acquisition module is electrically connected to the micro single-board computer, the micro single-board computer is electrically connected to the display module, and the data acquisition module is composed of a capacitance proximity sensor and a distance sensor.

[0009] Preferably, the capacitance proximity sensor includes an electrode layer and a dielectric layer. The electrode layer is respectively an emitting electrode layer and a receiving electrode layer, the dielectric layer is a dielectric layer with microstructures, the capacitance proximity sensor is a sandwich structure, and a packaging layer is arranged outside the electrode layer.

[0010] A capacitance proximity sensing substance recognition method based on machine learning includes: S1. Based on the sandwich structure with microstructures, prepare a capacitance proximity sensor, and attach a distance sensor to the capacitance proximity sensor to jointly form a data acquisition module; S2. Establish a machine learning model, collect target data through the data acquisition module, and train the machine learning model to evaluate and select the optimal machine learning model; S3. Build a substance recognition system. The micro single-board computer receives the digital signal collected by the data acquisition module, performs substance recognition according to the pre-established machine learning model, and sends the recognition result to the display module.

[0011] Preferably, the preparation of the capacitance proximity sensor in S1 includes: bonding the upper and lower electrode layers and the dielectric layer with microstructures based on the sandwich structure, and pasting a packaging layer after bonding.

[0012] Preferably, the microstructures in the dielectric layer with microstructures are prepared by printing a microstructure template with a 3D printer and using polydimethylsiloxane for casting replication to form polydimethylsiloxane with microstructures.

[0013] Preferably, the establishment of the machine learning model in S2 includes: S21. Collect target data. Attach a distance sensor to the capacitive proximity sensor to identify substances within the range of 15 mm - 25 mm. Place substances with different targets in sequence, detect the physical quantities of the substances and convert them into capacitance signals. S22. Data preprocessing. Extract the waveform feature data of the capacitance signals of the detected substances, normalize the data to obtain a data set, and divide the data set into a training set and a test set. The ratio of the training set to the test set is 8:2, which is used for the training and testing of the model. S23. Train and test the model. Select a suitable machine learning model according to the characteristics of the preprocessed data, use the training set data to train the machine learning model, and optimize the parameters of the machine learning model to achieve the optimal performance of the machine learning model. Use the test set to test and compare the trained model to select the optimal machine learning model for subsequent identification functions.

[0014] Therefore, the present invention adopts the above-mentioned capacitive proximity sensing substance recognition system and method based on machine learning, and the technical effects are as follows: 1. The capacitive proximity sensor does not need to be in direct contact with the substance, so it will not cause certain wear and irreversible deformation of the contact parts of the sensor, ensuring the stable quality of the signal transmitted by the sensor and enabling the sensor to collect substance signals more reliably.

[0015] 2. The capacitive proximity sensor based on machine learning can identify different substances, has strong applicability and versatility, and is not limited by the image quality and surface features of the substances compared with visual sensing recognition.

[0016] 3. A distance sensor is introduced to accurately define the detection range of the substance, significantly improving the measurement accuracy and reliability of the capacitive sensor.

[0017] 4. Compared with using a computer, the model transplanted to the micro single-board computer can significantly reduce the demand for hardware resources through lightweight optimization, and can meet the real-time ranging and substance recognition requirements in various scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the internal structure of the substance recognition system of the present invention; Figure 2 It is a schematic diagram of the recognition principle of the capacitive proximity sensor of the present invention; Figure 3 This is the working flow chart of the identification system of the present invention; Figure 4 It is a schematic diagram of the capacitance change amount of different substances between 15 - 25 mm from the distance sensor in the embodiment. Detailed implementation manners

[0019] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] Unless otherwise defined, the technical terms or scientific terms used in the present invention shall have the ordinary meanings understood by those of ordinary skill in the field to which the present invention belongs.

[0021] Embodiment 1 As Figure 1 shown, the present invention provides a capacitance proximity sensing substance identification system based on machine learning, including a system housing and an internal structure. The system housing consists of a display screen and a sensing area, and the internal structure consists of a power supply module, a data acquisition module, a micro single-board computer, and a display module. The power supply module is electrically connected to the micro single-board computer and the display module respectively, the data acquisition module is electrically connected to the micro single-board computer, the micro single-board computer is electrically connected to the display module, and the data acquisition module consists of a capacitance proximity sensor and a distance sensor.

[0022] When a substance approaches the system sensing area, the capacitance proximity sensor and the distance sensor work simultaneously to collect the proximity degree and distance information of the substance. After receiving this information, the micro single-board computer runs a pre-trained machine learning model for substance identification. The identification result is presented on the display screen through the display module, and the identified substance information can be visually seen. The introduction of the distance sensor enables the system to accurately measure the distance between the substance and the system sensing area, improving the measurement accuracy of the system.

[0023] As Figure 2 shown, the capacitance proximity sensor includes an electrode layer and a dielectric layer. The electrode layer is respectively an emission electrode layer and a receiving electrode layer, the dielectric layer is a dielectric layer with microstructures, the capacitance proximity sensor is a sandwich structure, and a packaging layer is arranged outside the electrode layer. When no substance approaches the sensor, the capacitance value between the two electrodes is small; when a substance approaches the sensor, the dielectric constant of the substance will affect the distribution of the electric field and the amount of charge on the sensing electrode of the sensor, resulting in an increase in the capacitance value between the two electrodes. By measuring the change in the capacitance value, the sensor can detect the proximity degree of the substance.

[0024] The principle of a capacitive proximity sensor for identifying substances is based on the edge electric field effect and the capacitance change mechanism. The sensor adopts a parallel plate electrode structure, and a voltage is applied between the two electrodes to form an electric field. Most of the electric field lines are parallelly distributed, and a small part diffuses outward to form an edge electric field. When no substance is close, the electric field distribution is stable; once a substance approaches, it will disturb the edge electric field distribution: on the one hand, an additional edge capacitance is formed between the substance and the electrode, and on the other hand, the substance will "steal" some charges of the driving electrode, resulting in a significant decrease in the mutual capacitance between the two electrodes. According to the capacitance definition formula , the decrease in the charge quantity Q directly causes the capacitance value to decrease, and the closer the substance is, the stronger the charge shunting effect and the greater the capacitance change. The sensor realizes non-contact sensing by detecting this capacitance change.

[0025] As Figure 3 shown, the present invention provides a method for identifying substances by capacitive proximity sensing based on machine learning. The specific steps are as follows: S1. Based on the sandwich structure of the micro-structure, a capacitive proximity sensor is prepared, and a distance sensor is attached to the capacitive proximity sensor to jointly form a data acquisition module; S2. A machine learning model is established. Target data is collected through the data acquisition module, and the machine learning model is trained. The optimal machine learning model is selected through evaluation; S3. A substance identification system is built. The micro single-board computer receives the digital signals collected by the data acquisition module, and performs substance identification according to the pre-established machine learning model, and sends the identification result to the display module.

[0026] During the process of preparing the capacitive proximity sensor, the upper and lower electrode layers and the dielectric layer with micro-structures are bonded based on the sandwich structure. After bonding, a packaging layer is pasted on. The micro-structures in the dielectric layer with micro-structures are prepared by printing a micro-structure template with a 3D printer and using polydimethylsiloxane for reverse mold replication to form polydimethylsiloxane with micro-structures.

[0027] The steps for establishing the machine learning model are as follows: S21. Target data is collected. A distance sensor is attached to the capacitive proximity sensor to identify substances within the range of 15 mm - 25 mm. Substances with different targets are placed in sequence, and the physical quantities of the substances are detected and converted into capacitance signals; S22. Data preprocessing. The waveform characteristic data of the capacitance signals of the detected substances is extracted, the data is normalized to obtain a data set, and the data set is divided into a training set and a test set. The ratio of the training set to the test set is 8:2, which is used for training and testing the model; S23. Train and test the model. Select a suitable machine learning model according to the characteristics of the preprocessed data. Use the training set data to train the machine learning model, and at the same time optimize the machine learning model parameters to achieve the optimal performance of the machine learning model. Use the test set to test and compare the trained model, and select the optimal machine learning model for subsequent identification functions.

[0028] As Figure 4 shown, in the experiment of non-contact substance identification, a VL53L1X laser range finder sensor module is attached to the sensor to achieve substance identification in the range of 15 mm - 25 mm. During the experiment, when the pressure plate drops to a distance of 25 mm from the capacitive sensor, the operating system automatically triggers the capacitive collector to start through a script, and only the capacitive data between 15 mm - 25 mm is collected during the collection process. Set the position of the sensor, and place substances with different targets, namely iron, wood block, leaf and polydimethylsiloxane, in turn, so that they are close to the capacitive sensor, and then detect the physical quantities of different substances and convert them into capacitive signals.

[0029] The principle of the capacitive proximity sensor to identify substances is that the dielectric constants of different materials will also affect the capacitance value. The dielectric constant of a leaf is usually between 3 and 5, the dielectric constant of PDMS is about 2.68, the dielectric constant of iron is usually considered to be infinite, the dielectric constant of a leaf is generally about 3, and the dielectric constant of a wood block is generally about 4.5. Furthermore, the dielectric constants of different substances affect the amplitude of the capacitance change. By comprehensively analyzing the capacitance changes caused by these factors, the capacitive non-contact sensor can identify and distinguish different substances.

[0030] During the operation of the substance identification system, the power supply module provides power supply for the data acquisition module, micro single-board computer and display module inside the system to keep the system operating normally; the distance sensor uses a VL53L1X laser range finder sensor to detect the distance between the substance and the sensor in real time, so that the substance can be in the precise detection range of the capacitive sensor; the capacitive proximity sensor is used to detect the approach of external substances. When a substance approaches the sensor, it will change the capacitance of the capacitor, and then the substance is identified by analyzing the change of the capacitance; the micro single-board computer, as the core of the whole system, receives the digital signal from the data acquisition card and performs substance identification according to the pre-established machine learning model, and at the same time sends the identification result to the display module; the display module receives the identification result from the micro single-board computer and displays the name and possible image of the substance on the screen.

[0031] The fusion technology of capacitive proximity sensors and VL53L1X laser range sensors can sense the spatial position of the target substance in real time and accurately define the substance range, thus significantly improving the performance of capacitive sensors. The application of model lightweight technology enables the trained machine learning model to be transplanted to embedded devices such as micro single-board computers, reducing the demand for hardware resources and achieving efficient operation in the ARM architecture and limited memory. This not only avoids the latency and privacy leakage risks of cloud data transmission but also supports local real-time processing of sensor data. Compared with the computer-side model that relies on cloud computing, the micro single-board computer has the advantages of low power consumption and low cost, and can meet the real-time ranging and substance identification requirements in various scenarios.

[0032] Therefore, the present invention adopts the above-mentioned capacitive proximity sensing substance identification system and method based on machine learning. By fusing capacitive proximity sensors and laser range sensors, high-precision substance identification is achieved. The model lightweight technology enables the system to operate efficiently on embedded devices, reducing costs, avoiding cloud latency and privacy leakage, and being applicable to various real-time ranging and substance identification scenarios.

[0033] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A capacitive proximity sensing material identification system based on machine learning, characterized in that: It includes a system shell and an internal structure, wherein the system shell is composed of a display screen and a sensing area, and the internal structure is composed of a power module, a data acquisition module, a micro single-board computer and a display module, wherein the power module is electrically connected to the micro single-board computer and the display module respectively, the data acquisition module is electrically connected to the micro single-board computer, the micro single-board computer is electrically connected to the display module, and the data acquisition module is composed of a capacitive proximity sensor and a distance sensor.

2. The capacitive proximity sensing material identification system based on machine learning according to claim 1, characterized in that: The capacitive proximity sensor comprises an electrode layer and a dielectric layer, wherein the electrode layers are respectively a transmitting electrode layer and a receiving electrode layer, and the dielectric layer is a dielectric layer with a microstructure. The capacitive proximity sensor is a sandwich structure, and a packaging layer is arranged outside the electrode layer.

3. A capacitive proximity sensing material identification method based on machine learning, characterized in that: include: S1. Based on the sandwich structure of the microstructure, a capacitive proximity sensor is prepared, and a distance sensor is attached to the capacitive proximity sensor to form a data acquisition module; S2. Establish a machine learning model, collect target data through the data acquisition module, train the machine learning model, and evaluate and select the optimal machine learning model; S3. Build a material identification system. The micro single-board computer receives the digital signal collected by the data acquisition module, and identifies the material according to the pre-established machine learning model, and sends the identification result to the display module.

4. The method for identifying substances by capacitive proximity sensing based on machine learning according to claim 3, characterized in that: The preparation of the capacitive proximity sensor in S1 includes: bonding the upper and lower electrode layers and the dielectric layer with a microstructure based on a sandwich structure, and attaching a packaging layer after bonding.

5. The method for identifying substances by capacitive proximity sensing based on machine learning according to claim 4, characterized in that: The microstructure in the dielectric layer with microstructure is prepared by printing a template with microstructure using a 3D printer, and using polydimethylsiloxane to perform reverse mold replication to form polydimethylsiloxane with microstructure.

6. The method for identifying substances by capacitive proximity sensing based on machine learning according to claim 3, characterized in that: The establishment of machine learning models in S2 includes: S21. Collect target data, attach a distance sensor to the capacitive proximity sensor, identify objects within a range of 15 mm to 25 mm, place different target materials in sequence, detect the physical quantity of the materials and convert them into capacitive signals; S22, data preprocessing, extracting the capacitance signal wave characteristic data of the detected substance, normalizing the data to obtain a data set, dividing the data set into a training set and a test set, with the ratio of the training set to the test set being 8:2, for model training and testing; S23. Training and testing models: Select a suitable machine learning model based on the characteristics of the preprocessed data, use the training set data to train the machine learning model, and optimize the machine learning model parameters to achieve the optimal machine learning model performance. Use the test set to test the trained model for comparative evaluation and select the optimal machine learning model for use in subsequent recognition functions.

Images