A method, device and computer readable medium for reading a dial gauge reading

Abstract

Some embodiments of the present disclosure provide a method, device and computer readable medium for identifying an analog instrument, the method comprising: inputting a first instrument panel image to be tested into a first neural network model, and outputting the first instrument panel image; extracting feature points of the first instrument panel image, and performing feature point registration on the first instrument panel and a standard instrument panel based on an image registration algorithm, and outputting a second instrument panel image after registration; inputting the second instrument panel image into a second neural network model, and outputting pointer position information on the second instrument panel image; and calculating a reading of the analog instrument based on the pointer position information. Embodiments of the present disclosure achieve identification and reading of the analog instrument based on an improved neural network model.

Description

Technical Field

[0001] The embodiments of this disclosure relate to the field of computer vision, specifically to a method, device, and computer-readable medium for recognizing readings on pointer-type instruments. Background Technology

[0002] To avoid the impact of electromagnetic interference on instruments, pointer-type instruments are still widely used in factories in industries such as power, petroleum, and coal. Currently, the existing methods for digitizing these pointer instruments typically involve manual inspection and recording, or using a fixed camera to capture images of a specific instrument and then recording them manually or through image recognition algorithms.

[0003] With the arrival of the 5G era and in response to the national strategic plan for intelligent manufacturing, inspection methods are constantly evolving towards informatization and intelligence. Currently, drones and robots are being used for inspections within factories. However, this method still has limitations. For example, due to weather or terrain limitations, manual inspection is still necessary when drones or robots cannot be used. Furthermore, with the development of augmented reality (AR) technology, manual inspection using AR devices (such as AR glasses) is becoming an indispensable method. The images for instrument recognition based on AR inspection technology are captured by cameras on wearable augmented reality display devices. Therefore, the captured instrument images are affected by the on-site environment, shooting angle, and how the device is worn, and existing instrument recognition methods cannot perfectly adapt to these factors. Summary of the Invention

[0004] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0005] Some embodiments of this disclosure provide a method for identifying pointer-type instrument readings and a computer-readable medium to address the technical problems mentioned in the background section above.

[0006] In a first aspect, some embodiments of this disclosure provide a method for recognizing the readings of a pointer-type instrument, comprising: inputting a test image having a first instrument panel into a first neural network model and outputting the first instrument panel image; extracting feature points from the first instrument panel image, registering the feature points of the first instrument panel and a standard instrument panel based on an image registration algorithm, and outputting a registered second instrument panel image; inputting the second instrument panel image into a second neural network model and outputting pointer position information on the second instrument panel image; and calculating the reading of the pointer-type instrument based on the pointer position information.

[0007] Secondly, some embodiments of this disclosure provide a device for recognizing a pointer-type instrument, comprising: a camera having first camera parameters and configured to capture a test image of a first instrument panel; an instrument panel recognition and positioning module including a first neural network model configured to input the test image and output the first instrument panel image; an image registration module configured to extract feature points from the first instrument panel image, register the first instrument panel and a standard instrument panel based on an image registration algorithm, and output a registered second instrument panel image; a pointer recognition and positioning module including a second neural network model configured to input the second instrument panel image and output pointer position information on the second instrument panel image; and a reading acquisition module configured to calculate the reading of the pointer-type instrument based on the pointer position information.

[0008] Thirdly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0009] The above-described embodiments of this disclosure have the following beneficial effects: Based on the basic module of separable convolution, a neural network model that can run in real time on low computing power devices is designed, enabling the algorithm to meet the needs of edge computing; a feature point matching algorithm is used to simultaneously perform dashboard positioning and pose correction to solve the problem of instrument reading error. Attached Figure Description

[0010] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0011] Figure 1This is a flowchart of a method for identifying pointer-type instrument readings according to one or more embodiments of the present disclosure;

[0012] Figure 2 This is a schematic diagram of the output of a first neural network structure according to one or more embodiments of the present disclosure;

[0013] Figure 3 This is a schematic diagram of a pointer reading calculation method according to one or more embodiments of the present disclosure;

[0014] Figure 4 This is a schematic diagram of a method for correcting the reading of a pointer instrument according to one or more embodiments of the present disclosure;

[0015] Figure 5 This is a schematic block diagram of a method for identifying pointer-type instrument readings according to one or more embodiments of the present disclosure. Detailed Implementation

[0016] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0017] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0018] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0019] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0020] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] Figure 1 A flowchart illustrating the pointer-type instrument reading recognition method applicable to this disclosure is shown. Specifically, it includes execution steps S1-S4:

[0022] S1. Input the image to be tested with the first dashboard into the first neural network model, and output the image of the first dashboard;

[0023] S2. Extract feature points from the first dashboard image, and based on the image registration algorithm, register the feature points of the first dashboard and the standard dashboard, and output the registered second dashboard image.

[0024] S3. Input the second dashboard image into the second neural network model and output the pointer position information on the second dashboard image;

[0025] S4. Calculate the reading of the pointer instrument based on the pointer position information.

[0026] Specifically, in step S1, the image to be tested may contain one or more pointer dashboards. The first neural network model can be a neural network model trained on an image of a pointer dashboard with its location marked. The first neural network model is configured to, after reading in the image to be tested, output the area range of the pointer dashboard in the image to be tested if the image contains a pointer dashboard. In one embodiment, such as... Figure 2 As shown, the image under test includes a pointer-type instrument panel and other interfering components. After passing through the first neural network model, the region of the pointer-type instrument panel in the image under test is output, and this region can be displayed as a rectangle. Subsequently, the region of the pointer-type instrument panel is cropped out by an image cropping method, and can be used as input data for subsequent image processing steps.

[0027] In one or more embodiments, the first neural network model can be a deep separable convolutional model. The deep separable convolutional model utilizes a factorization convolution operator instead of the full convolution operator. Factorization divides the standard convolution into two sub-convolutions: the first layer is a deep convolution that performs lightweight filtering on a single channel of the input data; the second layer is a pointwise convolution that computes the characteristics of the input channels to construct new feature vectors. Compared to the standard convolutional model, the deep separable convolutional model can significantly reduce computation without sacrificing accuracy. In one specific embodiment, the MobileNetV2 network architecture can be used as the deep convolutional model. In a preferred embodiment, the first neural network model can use the MobileNetV2 network architecture as the backbone network and the YOLOv5 model as the head network. In this preferred embodiment, due to the combination of the lightweight models of MobileNetV2 and YOLOv5, computation can be reduced while maintaining performance, making it suitable for use on mobile devices with moderate performance.

[0028] In step S2, feature points of the first dashboard image are extracted. Based on an image registration algorithm, the first dashboard and a standard dashboard are registered using feature points, and the registered second dashboard image is output. The feature points of the first dashboard image can be extracted using algorithms based on SIFT, SURF, ORB, etc. For example, the SIFT (Scale Invariant Feature Transform) algorithm can be used to extract and describe features. This description is scale invariant and can detect key points in the image; it is a local feature descriptor. Based on the SIFT feature descriptor, the image of the dashboard to be tested and the standard image can be matched to confirm whether the first dashboard to be tested matches one of the dashboards in the standard dashboard library. After confirming the standard dashboard that matches the first dashboard to be tested, the first dashboard and the standard dashboard can be registered using an image registration algorithm, and the registered second dashboard image is output. Image registration is the process of matching and superimposing two or more images acquired at different times, by different sensors (imaging devices), or under different conditions (weather, illumination, camera position and angle, etc.). Specifically, the first dashboard image can be converted into a second dashboard image based on affine transformation. The second dashboard image is an image with the same pose relationship as the standard image, so as to eliminate the influence of the second dashboard image's tilt, rotation, and other poses, thereby facilitating subsequent pointer reading processing.

[0029] In step S3, the second dashboard image is input into the second neural network model, which outputs pointer position information on the second dashboard image. Specifically, the second neural network model is a neural network model trained on an image of a pointer-type dashboard containing labeled pointer position information. In one or more embodiments, the second neural network model is a semantic segmentation network model. Semantic segmentation of an image involves assigning a semantic category to each pixel in the input image to obtain pixelated dense classification. A typical semantic segmentation architecture can be considered as an encoder-decoder network. The encoder is typically a pre-trained classification network, such as VGG or ResNet, followed by a decoder network. The main difference between these architectures lies in the decoder network. The decoder's task is to semantically map the discriminative features (lower resolution) learned by the encoder to the pixel space (higher resolution) to obtain dense classification. In one specific embodiment, the second neural network model can use a lightweight real-time semantic segmentation network model, such as the BiSeNet V2 model or the Unet model. Semantic segmentation is based on low-level details and high-level semantics. BiSeNet V2 is designed to process these spatial details and classification semantics separately to achieve high-precision and high-efficiency real-time semantic segmentation, balancing speed and accuracy. Due to its narrow channel and fast downsampling strategy, the computational cost of semantic branches in the BiSeNetV2 framework is also very low. In one embodiment, the pointer position information is pointer mask information. In another embodiment, the pointer position information is the coordinates of pointer keypoints, such as the coordinates of the needle tip, root, or center.

[0030] Specifically, in step S4, the reading of the pointer instrument is calculated based on the pointer position information. There are two main calculation methods for this step: one is the angle method, which calculates the angle between the line connecting the scale and the center of the circle and the pointer line, and uses the angle value to estimate the reading; the other is the distance method, which unfolds the circular image along the scale line arc to obtain a bar-shaped binary image, and estimates the reading using the pointer's projection coordinates on the horizontal axis. For example... Figure 3 As illustrated in the example, the specific calculation formula for the angle method is shown below.

[0031]

[0032] Where R min α is the reading of the starting scale of the dial, β is the angle between the starting scale and the pointer, and Range is the dial's range. The parameters necessary for the above formula calculation can be determined using information from the standard instrument panel matched in step S2.

[0033] Optionally, a preprocessing step for the image to be tested may be included before step S1. Image preprocessing may include one or more operations such as compression, cropping, erosion, dilation, grayscale conversion, and filtering to reduce interference from irrelevant information in the image and improve the efficiency of subsequent image processing. In one embodiment, a circular detection algorithm can be used to first detect a circle in the image to be tested, and the surrounding area of ​​the detected circle can be taken as the region of interest and cropped. Subsequently, the resolution of the cropped image can be adjusted to obtain image data that can be input to the next step. In another embodiment, an erosion-dilation operation can be performed on the image to be tested to reduce external noise; or a dilation-erosion operation can be performed on the image to be tested to reduce internal noise and reduce interference information in the first neural network recognition process. Of course, the above image preprocessing steps can be adjusted according to the different environments in which the image to be tested is located, thereby improving the overall running efficiency of the algorithm.

[0034] Optionally, in one embodiment, the image to be tested is captured by a camera with first camera parameters; based on the first camera parameters, a first positional relationship between the camera and the instrument panel in the captured image of the instrument to be tested is calculated; based on the first positional relationship and a known second positional relationship between the instrument panel and the pointer in the standard instrument panel, the reading error of the pointer instrument is calculated; and based on the reading error, the reading of the pointer instrument is corrected. Specifically, the camera parameters are inherent parameters of the camera and can be represented by a camera parameter matrix, which is usually fixed after the camera leaves the factory. The function of the camera parameters is to convert the image from the camera coordinate system to the imaging coordinate system. Therefore, when the camera parameters are known, the positional relationship between the captured object and the camera can be calculated from the two-dimensional image based on the pinhole model principle, where the positional relationship can include information such as shooting distance and shooting angle. After calculating the spatial positional relationship between the captured instrument panel and the camera, and when the spatial positional relationship between the instrument panel and the pointer is known, the reading error of the pointer instrument can be calculated. For example, as... Figure 4 As shown, a simplified model for calculating the reading error of a pointer-type instrument is illustrated. Figure 4 In the left image, the circular area represents the instrument panel. The solid arrow represents the actual position of the pointer, while the dashed arrow represents the position observed from the viewing angle. Because the plane on which the pointer is located is not the same as the plane on which the instrument panel scale is located, there is an error Δy. Δy can be calculated using... Figure 4 The geometric relationship calculation in the right figure can be performed based on the distance from the instrument to the camera calculated from the camera parameters, and the shooting angle θ can be calculated based on the affine transformation information in step S2. Then, Δy can be calculated using d*tanθ. This is understandable. Figure 4 The model is a simplified model; those skilled in the art can obtain a more accurate Δy based on a more complex model, but based on... Figure 4 The model already meets the requirements of the embodiments of the present invention.

[0035] Optionally, in an alternative embodiment, a third neural network model can be used to output the corrected instrument reading based on the pointer position information output in step S3. The third neural network model is a corrected reading regression model, which can employ a CNN neural network model with a certain depth. The output of the neural network model is modified to a single neuron, and the sigmoid function is used to limit the output value to the range of the instrument reading (e.g., 0-100), thus achieving the effect of pointer reading regression. Using the pointer position information output in step S3 as input and the corrected instrument reading as the label for model training, a trained model can be obtained. In one or more embodiments, the CNN model with a certain depth can be a ResNet18 model with 18 layers or a ResNet34 model with 34 layers. With the corrected reading regression model, since the label during training is already the corrected reading, the corrected instrument reading can be directly output, resulting in high accuracy. However, this method may increase the training workload due to the need to collect instrument data from different angles.

[0036] Optionally, in an alternative embodiment, the correction of instrument reading errors can also be implemented using a fourth neural network model. This fourth neural network model can be trained using images labeled with real instrument reading errors. It can be constructed using the ShuffleNetV2 model and can directly output the error value to be corrected based on the input image of the instrument under test. In this embodiment, the entire instrument identification process can be implemented using a neural network model, reducing the complexity of algorithm deployment and enhancing the adaptability of the entire process.

[0037] Based on steps S1-S4 and other optional steps, the positioning and reading recognition of pointer-type instrument panels can be achieved. Furthermore, the method of this invention is more applicable to instrument recognition based on non-fixed cameras, reducing reading errors caused by shooting at different angles and distances, resulting in more accurate final readings. This method is particularly important in AR inspections in industries such as power and petroleum, enabling efficient instrument recognition and effectively reducing the risk of false alarms and missed alarms. In addition, the neural network models in one or more embodiments of this invention employ lightweight network design, enabling offline use on lower-performance computing platforms (e.g., ARM-based mobile devices), greatly enhancing the applicability of this method in industrial scenarios.

[0038] like Figure 5 As shown, one or more embodiments of the present invention also disclose a device for recognizing a pointer-type instrument, comprising: a camera having first camera parameters and configured to capture a test image having a first instrument panel; an instrument panel recognition and positioning module including a first neural network model configured to: input the test image and output the first instrument panel image; an image registration module configured to: extract feature points from the first instrument panel image, register the feature points of the first instrument panel and a standard instrument panel based on an image registration algorithm, and output a registered second instrument panel image; a pointer recognition and positioning module including a second neural network model configured to: input the second instrument panel image and output pointer position information on the second instrument panel image; and a reading acquisition module configured to: calculate the reading of the pointer-type instrument based on the pointer position information.

[0039] like Figure 5 As shown by the dashed line, the pointer-type instrument recognition device may further include a preprocessing module. This preprocessing module can perform one or more operations such as compression, cropping, erosion, dilation, grayscale conversion, and filtering to reduce interference from irrelevant information in the image and improve the efficiency of subsequent image processing. The pointer-type instrument recognition device may also include an instrument reading correction module. This module can calculate the reading error based on the shooting distance and angle as described above, or by using a neural network algorithm, thereby improving the accuracy of the final algorithm. Each module of the instrument recognition device is configured to implement the specific steps described above.

[0040] In particular, according to some embodiments of this disclosure, the processes described above can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, it performs the functions defined in the methods of some embodiments of this disclosure.

[0041] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0042] The aforementioned computer-readable medium can be stored on smart devices such as mobile phones, watches, tablets, and glasses. The aforementioned computer-readable medium carries one or more programs that, when executed by the smart device, cause the smart device to: input a test image having a first dashboard into a first neural network model and output the first dashboard image; extract feature points from the first dashboard image, and based on an image registration algorithm, register the feature points of the first dashboard and a standard dashboard, outputting a registered second dashboard image; input the second dashboard image into a second neural network model and output pointer position information on the second dashboard image; and calculate the reading of the pointer-type instrument based on the pointer position information.

[0043] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0044] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0045] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0046] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for identifying the readings of a pointer-type instrument, comprising: The image to be tested with the first dashboard is input into the first neural network model, and the first dashboard image is output. The first neural network model is a depthwise separable convolutional model. The first neural network model uses the MobileNetV2 network structure as the backbone network and the YOLOV5 model as the head network. Feature points are extracted from the first dashboard image. Based on the image registration algorithm, the first dashboard and the standard dashboard are registered by feature points, and the registered second dashboard image is output. The second dashboard image is input into the second neural network model, and the pointer position information on the second dashboard image is output. The second neural network model is a semantic segmentation network model. The pointer position information is input into the third neural network, and the corrected reading of the pointer instrument is output. The third neural network model is a corrected reading regression model. The image to be tested was captured by a camera with first camera parameters; Based on the first camera parameters, calculate the first positional relationship between the camera and the instrument panel in the captured image of the instrument under test; Based on the first positional relationship and the second positional relationship between the instrument panel and the pointer known in the standard instrument panel, the reading error of the pointer-type instrument is calculated. Based on the reading error, the reading of the pointer instrument is corrected; Establish a standard instrument panel library, wherein the instrument information in the standard instrument panel library includes the structural model of the pointer and instrument panel, shooting distance, and shooting angle; The instrument information in the standard instrument panel library is captured by the same camera that captures the image under test.

2. The method according to claim 1, wherein calculating the reading error of the pointer-type instrument based on the first positional relationship and the second positional relationship between the instrument panel and the pointer known in the standard instrument panel further comprises: The standard instrument image is captured by the camera at a first shooting distance and a first angle. Based on the morphological relationship between the first dashboard image and the standard dashboard image, the second shooting distance and the second angle for capturing the image under test are calculated according to the first shooting distance and the first angle. The reading error of the pointer instrument is calculated based on the second shooting distance and the second angle.

3. A recognition device for a pointer-type instrument, comprising: A camera, having first camera parameters, is configured to capture a test image having a first instrument panel. Based on the first camera parameters, a first positional relationship is calculated between the camera and the instrument panel in the captured test image. Based on the first positional relationship and the second positional relationship between the instrument panel and the pointer known in the standard instrument panel, the reading error of the pointer instrument is calculated, and the reading of the pointer instrument is corrected based on the reading error; A standard instrument panel library is established, and the instrument information in the standard instrument panel library includes the structural model of the pointer and instrument panel, the shooting distance, and the shooting angle; the instrument information in the standard instrument panel library is captured by the same camera that captures the image to be tested. The dashboard recognition and positioning module includes a first neural network model, which is configured to: input the image to be tested and output the first dashboard image. The first neural network model is a depthwise separable convolutional model, which uses the MobileNetV2 network structure as the backbone network and the YOLOV5 model as the head network. The image registration module is configured to: extract feature points from the first dashboard image, register the feature points of the first dashboard and the standard dashboard based on the image registration algorithm, and output the registered second dashboard image; The pointer recognition and positioning module includes a second neural network model, which is configured to: input the second dashboard image and output pointer position information on the second dashboard image. The second neural network model is a semantic segmentation network model. The reading acquisition module is configured to input the pointer position information into a third neural network and output the corrected reading of the pointer instrument. The third neural network model is a corrected reading regression model.

4. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-2.

Images