Pointer instrument identification method, device and medium based on improved UNet

Abstract

This invention discloses a method, device, and medium for identifying pointer-type instruments based on an improved UNet, relating to the field of rail transit safety technology. The method includes: constructing an instrument panel dataset and training an improved UNet network using the dataset to obtain an instrument panel recognition model; acquiring real-time instrument panel images and inputting these images into the instrument panel recognition model to calculate instrument readings. This invention utilizes historical datasets to train the improved UNet network to obtain the instrument panel recognition model. Based on this model, key information in videos can be identified promptly, thereby quickly detecting abnormal readings of pointer-type instruments and minimizing potential safety hazards.

Description

Technical Field

[0001] This invention relates to the field of rail transit safety technology, and more specifically, to a pointer-type instrument identification method, device, and medium based on an improved UNet. Background Art

[0002] Pump rooms and other related facilities in high-speed railway station areas provide a range of services to the railway lines, stations, and production units, including flood control, drainage, fire fighting, and daily production and domestic water supply. These services have a direct impact on railway safety, train operation safety, and the quality of passenger services at the station, thus carrying significant responsibility.

[0003] Currently, the operation and maintenance of water supply and drainage systems on high-speed railways relies heavily on manpower, resulting in a relatively extensive management approach. A significant amount of human resources is devoted to manual duty at pump stations, regular maintenance, and line patrols. In the event of an emergency, rapid repairs are essential, requiring each station to have on-duty personnel to monitor equipment operation, leading to an enormous workload for maintenance services. Although existing pump stations are equipped with high-definition cameras to monitor the operating environment, the massive amounts of real-time video footage provided by these cameras lack automatic analysis capabilities. This makes it impossible to promptly identify key information in the video or quickly detect abnormal readings on pointer-type instruments, thus posing significant safety hazards. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a pointer-type instrument recognition method, device, and medium based on an improved UNet, which can quickly and accurately identify key information in videos and improve security.

[0005] In a first aspect, the present invention provides a pointer-type meter recognition method based on an improved UNet, the method comprising:

[0006] Construct a dashboard dataset and use the dashboard dataset to train an improved UNet network to obtain a dashboard recognition model;

[0007] Acquire a real-time dashboard image, input the real-time dashboard image into the dashboard recognition model, and calculate the dashboard values.

[0008] Furthermore, the improved UNet network includes a decoding layer and an encoding layer, wherein the encoding layer comprises 12 sequentially connected structural layers, wherein:

[0009] The first layer is a convolutional layer with a kernel of 3×3, a stride of 2, and 32 output channels.

[0010] The second layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 32 output channels.

[0011] The third layer is a pointwise convolutional layer with a 1×1 kernel, a stride of 1, and 64 output channels.

[0012] The fourth layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 64 output channels.

[0013] The 5th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 128 output channels.

[0014] The 6th layer is a deep convolutional layer with a kernel of 3×3, a stride of 2, and 128 output channels.

[0015] The 7th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 256 output channels.

[0016] The 8th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 256 output channels.

[0017] The 9th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 512 output channels.

[0018] The 10th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 512 output channels.

[0019] The 11th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 1024 output channels.

[0020] The 12th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 1024 output channels.

[0021] The results from layers 4, 6, 8, 10, and 12 serve as the input to the decoding layer of the UNet network.

[0022] Furthermore, an improved UNet network is trained using the aforementioned dashboard dataset to obtain a dashboard recognition model, including:

[0023] A grayscale mask is used as the dashboard dataset; wherein, the value of the large scale in the grayscale mask is 1, the value of the pointer is 2, and the rest are 0;

[0024] The dashboard dataset is imported into the improved UNet network and trained using the stochastic optimization method with adaptive momentum to obtain the dashboard recognition model.

[0025] Further, acquiring a real-time dashboard image, inputting the real-time dashboard image into the dashboard recognition model, and calculating the dashboard values ​​include:

[0026] Using a real-time dashboard image as the original image, a region of interest is set, and the original image is cropped to obtain a rectangular target image; the rectangular target image is then input into the dashboard recognition model to obtain a rectangular mask image M;

[0027] The rectangular mask M is expanded around its center point (c,c) to obtain a feature map T with width and height w and c respectively. The pixel values ​​of feature map T satisfy T(i,j)=M(m,n); where i and j are pixel coordinate indices, satisfying...

[0028]

[0029] In the formula, m is the x-coordinate of the image after unfolding, n is the y-coordinate of the image after unfolding, c is the x-coordinate of the center point of the mask image, ρ is the polar radius of the polar coordinate system with the center point as the origin, and θ is the polar angle of the polar coordinate system with the center point as the origin.

[0030] Determine the instrument scale index k q q = 1, 2, ..., n, where n represents the number of large scale divisions, satisfying 'a' represents the scale height threshold; 'k' determines the left edge index of the pointer. -1 ,satisfy b is the pointer height threshold; determine the right edge index k of the pointer. -2 ,satisfy and The pointer index is k0 = (k -1 +k -2 ) / 2;

[0031] According to k r <k0<k r+1 Determine the index r, where k r Let k be the index of the r-th instrument scale. r+1 This is the index of the (r+1)th instrument scale; calculate the current instrument value based on the preset range and index r.

[0032] Furthermore, based on the preset range and index r, the current instrument value is calculated using the following formula:

[0033]

[0034] In the formula, d is the instrument value and W is the preset range.

[0035] Secondly, the present invention provides a pointer-type meter recognition device based on an improved UNet, the device comprising:

[0036] The model training module is configured to construct a dashboard dataset and train an improved UNet network using the dashboard dataset to obtain a dashboard recognition model.

[0037] The instrument recognition module is configured to acquire real-time instrument panel images, input the real-time instrument panel images into the instrument panel recognition model, and calculate instrument values.

[0038] Furthermore, the improved UNet network includes a decoding layer and an encoding layer, wherein the encoding layer comprises 12 sequentially connected structural layers, wherein:

[0039] The first layer is a convolutional layer with a kernel of 3×3, a stride of 2, and 32 output channels.

[0040] The second layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 32 output channels.

[0041] The third layer is a pointwise convolutional layer with a 1×1 kernel, a stride of 1, and 64 output channels.

[0042] The fourth layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 64 output channels.

[0043] The 5th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 128 output channels.

[0044] The 6th layer is a deep convolutional layer with a kernel of 3×3, a stride of 2, and 128 output channels.

[0045] The 7th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 256 output channels.

[0046] The 8th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 256 output channels.

[0047] The 9th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 512 output channels.

[0048] The 10th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 512 output channels.

[0049] The 11th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 1024 output channels.

[0050] The 12th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 1024 output channels.

[0051] The results from layers 4, 6, 8, 10, and 12 serve as the input to the decoding layer of the UNet network.

[0052] Furthermore, the instrument identification module is further configured as follows:

[0053] Using a real-time dashboard image as the original image, a region of interest is set, and the original image is cropped to obtain a rectangular target image; the rectangular target image is then input into the dashboard recognition model to obtain a rectangular mask image M;

[0054] The rectangular mask M is expanded around its center point (c,c) to obtain a feature map T with width and height w and c respectively. The pixel values ​​of feature map T satisfy T(i,j)=M(m,n); where i and j are pixel coordinate indices, satisfying...

[0055]

[0056] In the formula, m is the x-coordinate of the image after unfolding, n is the y-coordinate of the image after unfolding, c is the x-coordinate of the center point of the mask image, ρ is the polar radius of the polar coordinate system with the center point as the origin, and θ is the polar angle of the polar coordinate system with the center point as the origin.

[0057] Determine the instrument scale index k q q = 1, 2, ..., n, where n represents the number of large scale divisions, satisfying and 'a' represents the scale height threshold; 'k' determines the left edge index of the pointer. -1 ,satisfy and b is the pointer height threshold; determine the right edge index k of the pointer. -2 ,satisfy and The pointer index is k0 = (k -1 +k -2 ) / 2;

[0058] According to k r <k0<k r+1 Determine the index r, where k r Let k be the index of the r-th instrument scale. r+1 This is the index of the (r+1)th instrument scale; calculate the current instrument value based on the preset range and index r.

[0059] Furthermore, the instrument identification module is further configured to calculate the current instrument value based on a preset range and index r using the following formula:

[0060]

[0061] In the formula, d is the instrument value and W is the preset range.

[0062] Thirdly, the present invention provides a readable storage medium storing one or more programs that can be executed by one or more processors to implement the method described above.

[0063] The present invention has at least the following beneficial effects:

[0064] This invention uses historical datasets to train an improved UNet network to obtain a dashboard recognition model. Based on the dashboard recognition model, key information in videos can be identified in a timely manner, thereby quickly detecting abnormal readings of pointer instruments and eliminating potential safety hazards as much as possible. Attached Figure Description

[0065] Figure 1 An overall flowchart of a pointer-type meter recognition method based on an improved UNet according to an embodiment of the present invention is shown.

[0066] Figure 2 A prediction flowchart of a pointer-type meter recognition method based on an improved UNet according to an embodiment of the present invention is shown.

[0067] Figure 3 A structural diagram of a pointer-type meter identification device based on an improved UNet according to an embodiment of the present invention is shown. Detailed Implementation

[0068] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and specific examples, but this is not intended to limit the present invention. If there is no necessary sequential relationship between the various steps described herein, the order in which they are described as examples should not be considered a limitation. Those skilled in the art should understand that the order can be adjusted, as long as it does not disrupt the logical consistency between them and render the entire process impossible.

[0069] This invention provides a pointer-type meter recognition method based on an improved UNet, such as... Figure 1 The diagram shows a flowchart of a pointer-type meter recognition method based on an improved UNet. This method includes steps S10 to S20, which are described in detail below.

[0070] S10: Construct a dashboard dataset and use the dashboard dataset to train an improved UNet network to obtain a dashboard recognition model.

[0071] In some embodiments, step S10 is performed by steps S11 to S12.

[0072] Step S11: Improve the UNet network to obtain the improved UNet network, whose coding layer structure is as follows:

[0073] The first layer is a convolutional layer with a kernel of 3×3, a stride of 2, and 32 output channels.

[0074] The second layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 32 output channels.

[0075] The third layer is a pointwise convolutional layer with a 1×1 kernel, a stride of 1, and 64 output channels.

[0076] The fourth layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 64 output channels.

[0077] The 5th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 128 output channels.

[0078] The 6th layer is a deep convolutional layer with a kernel of 3×3, a stride of 2, and 128 output channels.

[0079] The 7th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 256 output channels.

[0080] The 8th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 256 output channels.

[0081] The 9th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 512 output channels.

[0082] The 10th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 512 output channels.

[0083] The 11th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 1024 output channels.

[0084] The 12th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 1024 output channels.

[0085] The results of layers 4, 6, 8, 10, and 12 are the inputs to the original decoding layer of the UNet network.

[0086] Step S12: Use a grayscale mask as the dashboard dataset. The values ​​at the large scale of the grayscale mask are 1, the values ​​at the pointer are 2, and the rest are 0. Import the dashboard dataset into the improved UNet network for training to obtain the dashboard recognition model. The training uses the stochastic optimization method of adaptive momentum.

[0087] S20: Acquire a real-time dashboard image, input the real-time dashboard image into the dashboard recognition model, and calculate the dashboard values.

[0088] In some embodiments, step S20 is performed by steps S21 to S24.

[0089] S21: Set the region of interest, crop the original image to obtain a rectangular target image; input the target image into the trained network to obtain a rectangular mask image M.

[0090] S22: Expand the rectangular mask M around the center point (c,c) to obtain a feature map T with width and height w and height respectively, where the pixel values ​​satisfy T(i,j)=M(m,n) and the pixel coordinate indices i and j satisfy:

[0091]

[0092] In the formula, m is the x-coordinate of the image after it is unfolded, n is the y-coordinate of the image after it is unfolded, c is the x-coordinate of the center point of the mask image, ρ is the polar radius of the polar coordinate system with the center point as the origin, and θ is the polar angle of the polar coordinate system with the center point as the origin.

[0093] S23: Determine the instrument scale index k q q = 1, 2, ..., n, where n represents the number of large scale divisions, satisfying and 'a' represents the scale height threshold; 'k' determines the left edge index of the pointer. -1 ,satisfy and b is the pointer height threshold; determine the right edge index k of the pointer. -2 ,satisfy and The pointer index is k0 = (k -1 +k -2 ) / 2.

[0094] S24: Predict the range as W; determine the index r, satisfying k r <k0<k r+1 ; Calculate the current instrument value

[0095] To facilitate understanding by those skilled in the art, a specific embodiment is provided to illustrate a pointer-type instrument recognition method based on an improved UNet, comprising the following steps: Step 1: Construct an instrument panel dataset and train the improved UNet network to obtain an instrument panel recognition model; Step 2: Input real-time instrument panel images into the instrument panel recognition model and calculate instrument values.

[0096] Specifically, the improved training of the UNet network in step 1 includes: in step 12, the dashboard dataset format consists of the original image and its corresponding grayscale label image, with a total of 7294 images used for training; the training uses the adaptive momentum method, with hyperparameters of a learning rate of 0.005, a first-order moment decay rate of 0.85, a second-order moment decay rate of 0.95, and a small constant of 10. -7 .

[0097] Specifically, if Figure 2 As shown, the calculation of the instrument value in step 2 includes:

[0098] Step 22: Expand the graph M around the center point (224, 224) to obtain a feature graph T with a width and height of 512 and 224 respectively;

[0099] Step 23: The scale height threshold is 20, and the pointer height threshold is 100; the indexes of the 9 instrument scales are 52, 97, 142, 187, 227, 268, 310, 352, and 398; the left edge index of the pointer is 68, the left edge index of the pointer is 77; and the pointer index is 72.

[0100] Step 24: Predicted range is 1.6; index r is 1; calculate the current instrument value.

[0101] This invention also provides a pointer-type instrument recognition device based on an improved UNet, such as... Figure 3 As shown, the device includes:

[0102] The model training module 301 is configured to construct a dashboard dataset and train an improved UNet network using the dashboard dataset to obtain a dashboard recognition model.

[0103] The instrument recognition module 302 is configured to acquire a real-time instrument panel image, input the real-time instrument panel image into the instrument panel recognition model, and calculate the instrument values.

[0104] In some embodiments, the improved UNet network includes a decoding layer and an encoding layer, the encoding layer comprising 12 sequentially connected structural layers, wherein:

[0105] The first layer is a convolutional layer with a kernel of 3×3, a stride of 2, and 32 output channels.

[0106] The second layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 32 output channels.

[0107] The third layer is a pointwise convolutional layer with a 1×1 kernel, a stride of 1, and 64 output channels.

[0108] The fourth layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 64 output channels.

[0109] The 5th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 128 output channels.

[0110] The 6th layer is a deep convolutional layer with a kernel of 3×3, a stride of 2, and 128 output channels.

[0111] The 7th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 256 output channels.

[0112] The 8th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 256 output channels.

[0113] The 9th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 512 output channels.

[0114] The 10th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 512 output channels.

[0115] The 11th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 1024 output channels.

[0116] The 12th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 1024 output channels.

[0117] The results from layers 4, 6, 8, 10, and 12 serve as the input to the decoding layer of the UNet network.

[0118] In some embodiments, the instrument identification module is further configured to;

[0119] Using a real-time dashboard image as the original image, a region of interest is set, and the original image is cropped to obtain a rectangular target image; the rectangular target image is then input into the dashboard recognition model to obtain a rectangular mask image M;

[0120] The rectangular mask M is expanded around its center point (c,c) to obtain a feature map T with width and height w and c respectively. The pixel values ​​of feature map T satisfy T(i,j)=M(m,n); where i and j are pixel coordinate indices, satisfying...

[0121]

[0122] In the formula, m is the x-coordinate of the image after unfolding, n is the y-coordinate of the image after unfolding, c is the x-coordinate of the center point of the mask image, ρ is the polar radius of the polar coordinate system with the center point as the origin, and θ is the polar angle of the polar coordinate system with the center point as the origin.

[0123] Determine the instrument scale index k q q = 1, 2, ..., n, where n represents the number of large scale divisions, satisfying and 'a' represents the scale height threshold; 'k' determines the left edge index of the pointer. -1 ,satisfy and b is the pointer height threshold; determine the right edge index k of the pointer. -2 ,satisfy and The pointer index is k0 = (k -1 +k -2 ) / 2;

[0124] According to k r <k0<k r+1 Determine the index r, where k r Let k be the index of the r-th instrument scale. r+1 This is the index of the (r+1)th instrument scale; calculate the current instrument value based on the preset range and index r.

[0125] In some embodiments, the instrument identification module is further configured to calculate the current instrument value based on a preset range and index r using the following formula:

[0126]

[0127] In the formula, d is the instrument value and W is the preset range.

[0128] It should be noted that the structures of the various pointer-type instrument recognition devices based on the improved UNet described in this embodiment belong to the same technical concept as the pointer-type instrument recognition method based on the improved UNet described earlier, and achieve the same beneficial effects through the same principle, which will not be repeated here.

[0129] This invention also provides a readable storage medium storing one or more programs that can be executed by one or more processors to implement the methods described in any of the above embodiments.

[0130] Furthermore, although exemplary embodiments have been described herein, their scope includes any and all embodiments based on the invention that have equivalent elements, modifications, omissions, combinations (e.g., schemes involving intersections of various embodiments), adaptations, or alterations. Elements in the claims will be interpreted broadly based on the language used in the claims and are not limited to the examples described in this specification or during the implementation of this application, and such examples will be interpreted as non-exclusive. Therefore, this specification and examples are intended to be considered illustrative only, and the true scope and spirit are indicated by the following claims and the full scope of their equivalents.

[0131] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. Other embodiments can be used by those skilled in the art when reading the above description. Furthermore, in the above detailed description, various features may be grouped together to simplify the invention. This should not be construed as an intention that a feature of an unclaimed invention is necessary for any claim. Rather, the subject matter of the invention may be less than all the features of a particular embodiment of the invention. Thus, the following claims are incorporated herein by reference as examples or embodiments, wherein each claim is an independent, separate embodiment, and these embodiments are contemplated to be combined with each other in various combinations or arrangements. The scope of the invention should be determined by reference to the appended claims and the full scope of their equivalents.

Claims

1. A pointer-type instrument recognition method based on an improved UNet, characterized in that, The method includes: Construct a dashboard dataset and use the dashboard dataset to train an improved UNet network to obtain a dashboard recognition model; Acquire a real-time dashboard image, input the real-time dashboard image into the dashboard recognition model, and calculate the dashboard values; Acquire a real-time dashboard image, input the real-time dashboard image into the dashboard recognition model, and calculate the dashboard values, including: Using a real-time dashboard image as the original image, a region of interest is set, and the original image is cropped to obtain a rectangular target image; the rectangular target image is then input into the dashboard recognition model to obtain a rectangular mask image M; The rectangular mask M is expanded around its center point (c,c) to obtain a feature map T with width and height w and c respectively. The pixel values ​​of feature map T satisfy T(i,j)=M(m,n); where i and j are pixel coordinate indices, satisfying... In the formula, m is the x-coordinate of the image after unfolding, n is the y-coordinate of the image after unfolding, c is the x-coordinate of the center point of the mask image, ρ is the polar radius of the polar coordinate system with the center point as the origin, and θ is the polar angle of the polar coordinate system with the center point as the origin. Determine the instrument scale index k q q = 1, 2, ..., n, where n represents the number of large scale divisions, satisfying and 'a' represents the scale height threshold; 'k' determines the left edge index of the pointer. -1 ,satisfy and b is the pointer height threshold; determine the right edge index k of the pointer. -2 ,satisfy and The pointer index is k0 = (k -1 +k -2 ) / 2; According to k r <k0<k r+1 Determine the index r, where k r Let k be the index of the r-th instrument scale. r+1 This is the index of the (r+1)th instrument scale; calculate the current instrument value based on the preset range and index r.

2. The pointer-type instrument recognition method based on the improved UNet according to claim 1, characterized in that, The improved UNet network includes a decoding layer and an encoding layer. The encoding layer comprises 12 sequentially connected structural layers, wherein: The first layer is a convolutional layer with a kernel of 3×3, a stride of 2, and 32 output channels. The second layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 32 output channels. The third layer is a pointwise convolutional layer with a 1×1 kernel, a stride of 1, and 64 output channels. The fourth layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 64 output channels. The 5th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 128 output channels. The 6th layer is a deep convolutional layer with a kernel of 3×3, a stride of 2, and 128 output channels. The 7th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 256 output channels. The 8th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 256 output channels. The 9th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 512 output channels. The 10th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 512 output channels. The 11th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 1024 output channels. The 12th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 1024 output channels. The results from layers 4, 6, 8, 10, and 12 serve as the input to the decoding layer of the UNet network.

3. The pointer-type instrument recognition method based on the improved UNet according to claim 2, characterized in that, An improved UNet network is trained using the aforementioned dashboard dataset to obtain a dashboard recognition model, including: A grayscale mask is used as the dashboard dataset; wherein, the value of the large scale in the grayscale mask is 1, the value of the pointer is 2, and the rest are 0; The dashboard dataset is imported into the improved UNet network and trained using the stochastic optimization method with adaptive momentum to obtain the dashboard recognition model.

4. The pointer-type instrument recognition method based on the improved UNet according to claim 1, characterized in that, Based on the preset range and index r, the current instrument value is calculated using the following formula: In the formula, d is the instrument value and W is the preset range.

5. A pointer-type instrument recognition device based on an improved UNet, characterized in that, The device includes: The model training module is configured to construct a dashboard dataset and train an improved UNet network using the dashboard dataset to obtain a dashboard recognition model. The instrument recognition module is configured to acquire a real-time instrument panel image, input the real-time instrument panel image into the instrument panel recognition model, and calculate the instrument values; The instrument identification module is further configured as follows: Using a real-time dashboard image as the original image, a region of interest is set, and the original image is cropped to obtain a rectangular target image; the rectangular target image is then input into the dashboard recognition model to obtain a rectangular mask image M; The rectangular mask M is expanded around its center point (c,c) to obtain a feature map T with width and height w and c respectively. The pixel values ​​of feature map T satisfy T(i,j)=M(m,n); where i and j are pixel coordinate indices, satisfying... In the formula, m is the x-coordinate of the image after unfolding, n is the y-coordinate of the image after unfolding, c is the x-coordinate of the center point of the mask image, ρ is the polar radius of the polar coordinate system with the center point as the origin, and θ is the polar angle of the polar coordinate system with the center point as the origin. Determine the instrument scale index k q q = 1, 2, ..., n, where n represents the number of large scale divisions, satisfying and 'a' represents the scale height threshold; 'k' determines the left edge index of the pointer. -1 ,satisfy and b is the pointer height threshold; determine the right edge index k of the pointer. -2 ,satisfy and The pointer index is k0 = (k -1 +k -2 ) / 2; According to k r <k0<k r+1 Determine the index r, where k r Let k be the index of the r-th instrument scale. r+1 This is the index of the (r+1)th instrument scale; calculate the current instrument value based on the preset range and index r.

6. The pointer-type instrument recognition device based on the improved UNet according to claim 5, characterized in that, The improved UNet network includes a decoding layer and an encoding layer. The encoding layer comprises 12 sequentially connected structural layers, wherein: The first layer is a convolutional layer with a kernel of 3×3, a stride of 2, and 32 output channels. The second layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 32 output channels. The third layer is a pointwise convolutional layer with a 1×1 kernel, a stride of 1, and 64 output channels. The fourth layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 64 output channels. The 5th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 128 output channels. The 6th layer is a deep convolutional layer with a kernel of 3×3, a stride of 2, and 128 output channels. The 7th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 256 output channels. The 8th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 256 output channels. The 9th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 512 output channels. The 10th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 512 output channels. The 11th layer is a pointwise convolutional layer with a kernel of 1×1, a stride of 1, and 1024 output channels. The 12th layer is a deep convolutional layer with a kernel of 3×3, a stride of 1, and 1024 output channels. The results from layers 4, 6, 8, 10, and 12 serve as the input to the decoding layer of the UNet network.

7. The pointer-type instrument recognition device based on the improved UNet according to claim 6, characterized in that, The instrument identification module is further configured to calculate the current instrument value based on a preset range and index r using the following formula: In the formula, d is the instrument value and W is the preset range.

8. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, perform the method according to any one of claims 1 to 4.

Images