Object shape recognition system and method based on neural network and sensor array
By identifying object shapes under the combination of sensor arrays and neural networks, the problem of insufficient accuracy and robustness of traditional visual recognition methods in complex environments is solved, and more efficient and flexible object shape recognition is achieved.
Patent Information
- Application Number
- CN202411953871.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-27
- Publication Date
- 2025-05-27
AI Technical Summary
Traditional vision-based shape recognition methods have problems with identification accuracy and robustness in complex environments, especially for transparent or reflective objects that are difficult to effectively identify.
An object shape recognition system based on neural networks and sensor arrays is adopted to collect pressure signals at different points of the object through the sensor array, and the data is extracted and analyzed by using convolutional neural networks (CNNs) to identify the object shape.
In complex environments, various object shapes can be reliably identified, which improves the accuracy and adaptability of recognition, and overcomes the difficulties in identifying traditional methods under factors such as light changes, occlusions, and perspective changes.
Smart Images
Figure CN120045060A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of object shape recognition, and particularly to an object shape recognition system and method based on a neural network and a sensor array. Background Art
[0002] Shape recognition is an important research direction in the field of computer vision and is widely applied in multiple fields such as automated production, robot navigation, medical image analysis, etc. Traditional shape recognition mainly relies on visual sensors such as cameras. By capturing the appearance image of an object, image processing techniques are used to extract shape features, and finally pattern matching and classification are performed. This method generally includes several steps such as image acquisition, image preprocessing, feature extraction, and pattern matching classification. Image acquisition uses a camera to capture the image of an object. Image preprocessing performs operations such as denoising, grayscale conversion, and edge detection on the original image to enhance the image quality. Feature extraction extracts the shape features of the object through geometric shape descriptors (such as contours, corner points, edges, etc.). Finally, algorithms such as Template Matching, Shape Context, and Support Vector Machine (SVM) are used to compare the extracted features with known shapes in the database, thereby identifying the object.
[0003] However, shape recognition methods based on vision have some limitations. For example, factors such as illumination changes, occlusion, perspective transformation, and complex backgrounds will all affect the accuracy and robustness of recognition. In addition, for some transparent or reflective objects, traditional vision methods are difficult to effectively identify.
[0004] The object shape recognition technology based on a neural network and a sensor array combines advanced sensor technology and deep learning algorithms. Through multi-modal data fusion and intelligent analysis, it improves the accuracy and adaptability of shape recognition. A sensor array composed of multiple sensors is adopted to obtain the pressure information at different points when the hand grasps objects of different shapes. By analyzing this pressure information, objects of different shapes can be reflected.
[0005] The data from sensors at different points are fed into a Convolutional Neural Network (CNN), and the CNN extracts and analyzes the features of the data. The neural network has powerful feature learning and classification capabilities and can adaptively extract complex data features, significantly improving the accuracy of recognition.
[0006] By combining a neural network and a sensor array, the shape recognition system can more accurately and robustly recognize the shape of an object in a complex environment, which has important application value and broad prospects for fields such as industrial automation and intelligent robots. Summary of the Invention
[0007] The object of the present invention is to solve the technical problems existing in the background art. For this purpose, an object shape recognition system and method based on a neural network and a sensor array are provided.
[0008] In order to achieve the above object, the technical solutions adopted by the present invention are as follows:
[0009] An object shape recognition system based on a neural network and a sensor array, comprising:
[0010] A sensor array module for collecting pressure signals at different points when grasping an object.
[0011] A sensor data acquisition module for collecting and converting the pressure signals of the sensor array module, and amplifying the signals. The sensor data acquisition module is connected to the sensor array module.
[0012] A data sending module for sending the pressure signals collected by the sensor data acquisition module and sending them to the PC side via Bluetooth. The data sending module is connected to the sensor data acquisition module.
[0013] A neural network training module for extracting and analyzing the data collected by the sensor data acquisition module. The neural network training module is connected to the data sending module via Bluetooth.
[0014] The following is a further limited technical solution of the system in the present invention. The sensor array module is composed of nine sensors distributed on gloves or hand-shaped devices. The sensor array distributions are as follows: there is one sensor on the thumb, and there is one sensor on each of the fingers and the knuckles close to the palm of the index finger, middle finger, ring finger and little finger, for a total of nine sensors, so as to obtain data information at different points.
[0015] The following is a further limited technical solution of the system in the present invention. The sensor data acquisition module is composed of a resistor and an operational amplifier: the first operational amplifier realizes the conversion of current signals and voltage signals; the second operational amplifier amplifies and outputs the previous voltage signal, and an adjustable resistor is used in the circuit to realize the adjustment of the amplification factor.
[0016] The following is a further limited technical solution of the system in the present invention. The data sending module is implemented using STM32F104C8T6, and realizes the transmission of data via Bluetooth. The data sending module is connected to the sensor data acquisition module.
[0017] An object shape recognition method based on a neural network and a sensor array. The method is implemented based on the above object shape recognition system based on a neural network and a sensor array, and includes the following steps:
[0018] Step S1: Obtain the pressure information at different points when grasping an object through the sensor array module;
[0019] Step S2: Extract and amplify the signals of the sensor array module through the sensor data acquisition module;
[0020] Step S3: Send the data collected by the sensor data acquisition module to the PC side via Bluetooth through the data sending module;
[0021] Step S4: Extract and analyze the signal features through the neural network training module to distinguish different data.
[0022] The following is a further limited technical solution of the method in the present invention. The neural network training module uses a convolutional neural network (CNN) to train the collected data to achieve the distinction of different data. Among them, the dataset is divided according to the ratio of training set: validation set: test set = 6:1:1.
[0023] Compared with the prior art, the present invention has the following technical effects:
[0024] 1. Enhance the adaptability to complex environments: Traditional vision-based recognition methods have difficulties in recognizing complex backgrounds, transparent and reflective objects, external light changes, occlusion, perspective transformation and other factors. The present invention can reliably recognize various object shapes in complex environments and has stronger environmental adaptability.
[0025] 2. High efficiency: Through advanced neural network algorithms, the present invention can efficiently process and analyze multi-channel data to achieve accurate recognition and feedback of object shapes.
[0026] 3. Flexibility and scalability: The system architecture of the present invention has high flexibility, can flexibly configure different numbers of sensors according to specific application requirements, and also supports the upgrade and expansion of algorithms and hardware.
[0027] The present invention will be further described below in conjunction with the drawings and embodiments. Description of the Drawings
[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0029] Figure 1 It is a schematic diagram of the sensor array arrangement of the present invention;
[0030] Figure 2 It is a schematic diagram of the principle of the sensor data acquisition module of the present invention;
[0031] Figure 3 It is a schematic diagram of the pins of the sensor data acquisition module of the present invention;
[0032] Figure 4 It is a schematic diagram of the principle of the data sending module of the present invention;
[0033] Figure 5 It is a diagram of the module connection relationship of the present invention. Specific Embodiments
[0034] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.
[0035] As Figures 1-5 shown, this embodiment provides an object shape recognition system based on a neural network and a sensor array, which mainly consists of a sensor array module, a sensor data acquisition module, a data sending module, a neural network training module, etc.
[0036] As Figure 1 shown, the sensor array module is jointly composed of nine sensors distributed on gloves or hand-shaped devices, and is used to collect the pressure signals at different points during the grasping process. When grasping objects of different shapes, the force conditions at each point and the overall situation reflected by the array composed of the nine points will be different, thereby reflecting the appearance characteristics of the object.
[0037] As Figure 2As shown, the sensor data acquisition module consists of two operational amplifiers U1 and U2 (model LM358ADR). The inverting input terminal of operational amplifier U1 is electrically connected to one end of resistor R2, the other end of resistor R2 is electrically connected to one end of resistor R1, and the other end of resistor R1 is connected to the power supply voltage; the output terminal of sensor RX is electrically connected to the line between resistor R1 and resistor R2; the non-inverting input terminal of operational amplifier U1 is electrically connected to one end of resistor R3 and one end of resistor R5 respectively. Among them, the other end of resistor R5 is grounded, and the other end of resistor R3 is connected to the power supply voltage; a resistor R4 is electrically connected between the inverting input terminal and the output terminal of operational amplifier U1; the output terminal of operational amplifier U1 is electrically connected to the non-inverting input terminal of operational amplifier U2; the inverting input terminal of operational amplifier U2 is electrically connected to the movable end of adjustable resistor R6, one fixed end of adjustable resistor R6 is electrically connected to one end of resistor R7, the other end of resistor R7 is grounded, and the other fixed end of adjustable resistor R6 is electrically connected to the output terminal of operational amplifier U2.
[0038] Therefore, the first operational amplifier U1 realizes the conversion of the current signal into a voltage signal. When current flows through R1, a voltage difference will be generated across R1. According to the virtual open, it can be obtained that no current flows through the input terminal of the operational amplifier, so:
[0039] (V5 - V2) / R3 = V2 / R5 (1)
[0040] (V4 - V1) / R2 = (V1 - V3) / R4 (2)
[0041] According to the virtual short, it can be obtained that:
[0042] V1 = V2 (3)
[0043] Keeping R2 = R3; R4 = R5, from the above three equations, it can be obtained that
[0044]
[0045] Among them, (V5 - V4) represents the voltage difference across R1, that is, the converted voltage signal. To ensure that the output of the operational amplifier is not distorted, a conservative approach is taken, making R2 = R4, and finally V3 = (V5 - V4).
[0046] The second operational amplifier U2 mainly plays the role of amplifying the voltage signal. Similarly, according to the principle of virtual short and virtual open, it can be obtained that:
[0047] VOUT = V3(R6 + R7) / (X + R7) (5)
[0048] Among them, X represents the resistance value of the adjustable resistor R6 near the R7 part. In addition, when X is 0, R7 ensures that the inverting terminal of the operational amplifier is not directly grounded. RX represents the sensor.
[0049] As Figure 3 shown, the pin headers of the sensor data acquisition module are described. Among them, the two left pin headers represent the access terminals of the sensor, and the positive and negative poles do not need to be distinguished; the three right pin headers are power supply, signal output, and ground from top to bottom in sequence.
[0050] As Figure 4 shown, the signal sending end built on the STM32F104C8T6 platform, that is, the data sending module, is described. One end is connected to the sensor data acquisition module. The power supply and ground of the sensor data acquisition module are connected to the STM32F104C8T6, and the signal end is connected to the functional I / O pin of the STM32F104C8T6. Through the TXD / RXD serial communication pins in the STM32F104C8T6, it is connected to the Bluetooth, and data is transmitted through the Bluetooth to the PC to realize wireless data transmission.
[0051] As Figure 5 shown, the sensor array module is connected to the sensor data acquisition module, and the sensor data acquisition module is connected to the data sending module. Among them, the sensor data acquisition module and the data sending module share the power supply and ground.
[0052] The above is only a preferred embodiment of the present invention, and it does not impose any form of limitation on the present invention. Any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention, or modify it into an equivalent embodiment with equivalent changes, without departing from the scope of the technical solution of the present invention. Therefore, all equivalent changes made according to the shape, structure, and principle of the present invention without departing from the content of the technical solution of the present invention shall be covered by the protection scope of the present invention.
Claims
1. Object shape recognition system based on neural network and sensor array, characterized in that: include: A sensor array module is used to obtain information at different points when grasping an object; A sensor data acquisition module, connected to the sensor array module, for extracting and amplifying signals from the sensor array module; The data sending module sends the data collected by the sensor data acquisition module to the PC via Bluetooth; The neural network training module extracts and analyzes the features of the signals collected by the sensor data acquisition module through the neural network.
2. The object shape recognition system based on a neural network and a sensor array as claimed in claim 1, characterized in that: The sensor array module includes a plurality of sensors to obtain data information at different points.
3. The object shape recognition system based on neural network and sensor array as claimed in claim 1, characterized in that: The sensor data acquisition module includes a current-voltage conversion circuit and a voltage amplification circuit, wherein the current-voltage conversion circuit is used for converting current and voltage signals, and the voltage amplification circuit is used for amplifying voltage signals.
4. The object shape recognition system based on a neural network and a sensor array as claimed in claim 1, characterized in that: The data sending module is built based on STM32F104C8T6 and is connected to the sensor data acquisition module.
5. A method for object shape recognition based on a neural network and a sensor array, wherein the method is implemented based on the object shape recognition system based on a neural network and a sensor array according to any one of claims 1 to 4, characterized in that: The following steps are involved: Step S1: obtaining information of different points when grasping an object through a sensor array module; Step S2: extracting and amplifying the signal of the sensor array module through the sensor data acquisition module; Step S3: The data collected by the sensor data collection module is sent to the PC via Bluetooth by the data sending module; Step S4: Extract and analyze signal features through a neural network training module to distinguish different data.
6. The object shape recognition method based on a neural network and a sensor array as claimed in claim 5, characterized in that: The neural network training module uses a convolutional neural network to train the collected data, wherein the data set is divided according to the ratio of training set: validation set: test set = 6:1:
1.
7. The object shape recognition method based on neural network and sensor array as claimed in claim 5, characterized in that: The sensor array module is composed of N pressure sensors distributed on a glove or hand-shaped device, so as to collect pressure signals at N points when grasping an object, thereby analyzing the appearance characteristics of the object, where N≥9.