Reading method for reading of pointer instrument, and device, system and storage medium

Abstract

Provided in the embodiments of the present disclosure are a reading method for a reading of a pointer instrument, and a device, a system and a storage medium. In the present disclosure, scale number recognition is performed on a pointer instrument image, such that a scale number recognition result is obtained; by taking as prompt information image position information of a detection box for a pointer of a pointer instrument, pointer segmentation is performed on the pointer instrument image; and by taking as prompt information target image position information determined on the basis of a pointer segmentation result and the scale number recognition result, scale segmentation is performed on the pointer instrument image. Therefore, the pointer segmentation and the scale segmentation can be completed in the case of zero-shot learning, and semantic segmentation of the pointer instrument image can also be implemented without studying the pointer instrument corresponding to the pointer instrument image. Since there is no need to study a pointer instrument of a specific type, pointer segmentation can be performed on an image of a pointer instrument of an unknown type, thereby improving the universality of semantic segmentation of a pointer instrument image, and further improving the universality of the reading of a reading of the pointer instrument.

Description

[0001] Cross-Reference to Patent Application No. 202311601572.0, filed with the Chinese Patent Office on November 27, 2023, and entitled "Pointer Instrument Reading Method, Device, System, and Storage Medium," the entire contents of which are incorporated herein by reference. Technical Field This disclosure relates to the field of image processing technology, and more particularly to a pointer instrument reading method, device, system, and storage medium. Background: Pointer instruments are commonly used in industry, typically consisting of scale lines, scale numerals, and a pointer. In some applications, a large number of pointer instruments exist, and manual reading of these instruments is undoubtedly time-consuming and labor-intensive. To reduce the labor cost of reading pointer instruments, some traditional approaches capture a pointer instrument image and use object detection and semantic segmentation models to locate the pointer and scale in the pointer instrument image. The pointer instrument reading is then calculated using known range information and the processed pointer endpoint's relative position ratio to the scale. However, this pointer instrument reading method cannot read any type of pointer instrument image input, and its versatility and universality are low. SUMMARY OF THE INVENTION Various aspects of the present disclosure provide a pointer instrument reading method, device, system, and storage medium to improve the versatility of the pointer instrument reading method. An embodiment of the present disclosure provides a method for reading a pointer instrument, comprising: performing scale digit recognition on a pointer instrument image including an image of a dial of the pointer instrument to determine a scale digit recognition result of the pointer instrument; performing pointer detection on the pointer instrument image to determine image position information of a detection frame of a pointer of the pointer instrument; performing image segmentation on the pointer instrument image using the image position information of the pointer detection frame as prompt information to obtain a target pointer segmentation result; determining target image position information for scale segmentation prompting based on the target pointer segmentation result and the scale digit recognition result; performing image segmentation on the pointer instrument image using the target image position information as prompt information to obtain a target scale segmentation result; and determining a reading of the pointer instrument based on the scale digit recognition result, the target pointer segmentation result, and the target scale segmentation result.The present disclosure also provides a data center inspection system, comprising: an autonomous mobile device and a server device; the autonomous mobile device is provided with an image acquisition device; the autonomous mobile device is configured to move within a data center and, during movement, control the image acquisition device to acquire images of pointer instruments in the data center to obtain pointer instrument images including dial images of the pointer instruments; the pointer instrument images are provided to the server device; and the server device is configured to execute the steps of the pointer instrument reading method described above. The present disclosure also provides an electronic device, comprising: a memory and a processor; the memory is configured to store a computer program; the processor is coupled to the memory and configured to execute the computer program to execute the steps of the pointer instrument reading method described above. The present disclosure also provides a computer-readable storage medium storing computer instructions, which, when executed by one or more processors, causes the one or more processors to execute the steps of the pointer instrument reading method described above. In an embodiment of the present disclosure, scale digit recognition is performed on a pointer instrument image to obtain a scale digit recognition result. The pointer of the pointer instrument image is segmented using the image position information of the detection frame of the pointer of the pointer instrument as prompt information. Furthermore, the scale of the pointer instrument image is segmented using the target image position information determined based on the pointer segmentation result and the scale digit recognition result as prompt information. This allows for the segmentation of the pointer and scale without any learning required, without the need to learn the pointer instrument corresponding to the pointer instrument image, and also achieves semantic segmentation of the pointer instrument image. Because learning is not required for a specific pointer instrument type, pointer segmentation can be performed on images of unknown pointer instruments, improving the versatility and universality of semantic segmentation of pointer instrument images, and thus improving the versatility and universality of pointer instrument reading. BRIEF DESCRIPTION OF THE DRAWINGS The drawings described herein are provided to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The exemplary embodiments of the present disclosure and their description are provided for illustrative purposes only and are not intended to unduly limit the present disclosure.In the accompanying drawings: Figure 1a is a schematic diagram of the structure of a data center inspection system according to an embodiment of the present disclosure; Figure 1b is a schematic flow diagram of a pointer instrument reading method according to an embodiment of the present disclosure; Figure 1c is a schematic diagram of the specific process of the pointer instrument reading method according to an embodiment of the present disclosure; Figure 2 is a schematic diagram of the pointer segmentation process according to an embodiment of the present disclosure; Figure 3 is a schematic diagram of the pointer segmentation result according to an embodiment of the present disclosure; Figure 4 is a schematic diagram of the dial layout of a pointer instrument according to an embodiment of the present disclosure; Figure 5 is a schematic diagram of the scale segmentation result according to an embodiment of the present disclosure; and Figure 6 is a schematic diagram of the structure of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS To further clarify the objectives, technical solutions, and advantages of the present disclosure, the technical solutions of the present disclosure will be described clearly and completely below in conjunction with specific embodiments and corresponding drawings. It should be understood that the described embodiments are only some of the embodiments of the present disclosure, and are not exhaustive. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present disclosure without inventive effort are intended to fall within the scope of protection of the present disclosure. Traditional solutions for identifying pointer instruments based on pointer instrument images use object detection and semantic segmentation models to locate the pointer and scale in the pointer instrument image. The pointer instrument reading is then calculated using known range information and the processed position ratio of the pointer endpoint relative to the scale. These traditional pointer instrument reading solutions typically require training a semantic segmentation model for a specific pointer instrument. This requires labeling a large number of semantically segmented examples of pointer instrument images for that specific instrument. The labeled semantic segmentation examples are then used to train the semantic segmentation model, resulting in a semantic segmentation model specifically designed for that instrument. This trained semantic segmentation model can only perform semantic segmentation for the specific pointer instrument it has learned about, and cannot perform semantic segmentation for other unlearned types of pointer instruments. Consequently, the semantic segmentation results cannot be used to read pointer instruments of unlearned types. Consequently, these traditional pointer instrument reading solutions have limited versatility and generalizability. In some embodiments of the present disclosure, to improve the versatility and universality of pointer instrument reading methods, a solution is proposed. The basic concept is as follows: scale digit recognition is performed on a pointer instrument image to obtain a scale digit recognition result; pointer segmentation is performed on the pointer instrument image using the image position information of the pointer detection frame of the pointer instrument as prompt information; and scale segmentation is performed on the pointer instrument image using the target image position information determined based on the pointer segmentation result and the scale digit recognition result as prompt information. This solution can complete the segmentation of the pointer and scale without learning the pointer instrument corresponding to the pointer instrument image, and can also achieve semantic segmentation of the pointer instrument image.Since learning for specific types of pointer instruments is not required, pointer segmentation can be performed on images of unknown pointer instruments, improving the versatility and universality of semantic segmentation of pointer instrument images, and thus improving the versatility and universality of pointer instrument readings. The following, in conjunction with the accompanying drawings, describes in detail the technical solutions provided by various embodiments of the present disclosure. It should be noted that identical reference numerals in the following figures and embodiments represent identical objects. Therefore, once an object is defined in one figure or embodiment, it need not be further discussed in subsequent figures and embodiments. Figure 1a is a schematic diagram of the structure of a data center inspection system provided by an embodiment of the present disclosure. As shown in Figure 1b, the data center inspection system may include an autonomous mobile device 10 and a server device 20. Data centers have a variety of different pointer instruments configured to monitor various data center performance indicators. For example, data centers are equipped with pressure gauges configured to monitor the pressure of the data center's cooling system; ammeters and voltmeters configured to monitor the current and voltage of the data center's cabinets, etc. Data center maintenance personnel can understand the health status of the data center based on the monitoring results, i.e., readings, of these pointer instruments. To reduce manual training costs, autonomous mobile devices 10 are often used for data center inspections. In this embodiment, an autonomous mobile device 10 can move autonomously and complete certain tasks based on this autonomous movement. The specific implementation form of the autonomous mobile device 10 is not limited in this disclosed embodiment. The autonomous mobile device 10 can be implemented as a robot or drone, for example. The robot can have a humanoid, animal, vehicle, or puppet-like appearance. In this embodiment, as shown in FIG. 1a , an image acquisition device 11 is installed on the autonomous mobile device 10. In this embodiment, the image acquisition device can be implemented as any device with image acquisition capabilities, such as a camera, a still camera, or a video recorder. The images captured by the image acquisition device can be independent frames or video frames from a video. In this embodiment, the autonomous mobile device 10 can move within a data center. In this disclosed embodiment, the autonomous mobile device 10 refers to a device with an independent power system. The autonomous mobile device 10 can move using its own power system. The power system can include drive wheels, a drive motor, and a transmission device. The autonomous mobile device 10 can automatically move along an inspection route using its own power system. For example, the autonomous mobile device 10 can automatically plan an inspection route and automatically move along the route. Of course, the autonomous mobile device 10 can also be controlled by a user or other device to move in the data center. For example, a computing device (such as the server device 20) controls the autonomous mobile device 10 to move in the data center.Optionally, the computing device can send the inspection route to the autonomous mobile device 10 and control the autonomous mobile device 10 to move within the data center according to the inspection route. For another example, a user can control the autonomous mobile device 10 to move within the data center via a terminal such as a mobile phone or remote control. The autonomous mobile device 10 can respond to terminal control signals and utilize its own power system to move. Specifically, the autonomous mobile device 10 moves within the data center and, during movement, controls the image acquisition device 11 to capture images of the data center's pointer instruments, obtaining pointer instrument images including images of the pointer instrument dials. In this embodiment of the present disclosure, to reduce duplicate image acquisition, multiple acquisition locations can be pre-set along the inspection route. The pointer instrument images captured by the image acquisition device 11 at these multiple acquisition locations can cover all pointer instruments in the data center. In this embodiment of the present disclosure, the multiple acquisition locations can be determined based on the image acquisition device 11's acquisition angle of view and the distance between the inspection route and the pointer instruments. Of course, the image acquisition device can also manually test the acquisition range of the pointer instruments along the inspection route in advance. Based on the manual test results, multiple acquisition locations can be determined so that the pointer instrument images captured by the image acquisition device 11 at the multiple acquisition locations cover all the pointer instruments in the data center. Each acquisition location is configured to capture images of a portion of the pointer instruments in the data center. Based on the pre-set multiple acquisition locations, the autonomous mobile device 10 can autonomously locate itself while moving within the data center along the set inspection route to determine its current location. The specific implementation method for autonomous positioning by the autonomous mobile device 10 is not limited in the embodiments of the present disclosure. In some embodiments, the autonomous mobile device 10 can employ Simultaneous Localization and Mapping (SLAM) technology for autonomous positioning. Specifically, the autonomous mobile device 10 collects environmental information surrounding its current location and, based on this environmental information, locates its position within a stored environmental map. Optionally, the autonomous mobile device 10 may construct a temporary map based on the environmental information acquired during the movement process; and compare the constructed temporary map with the stored environmental map to determine the position and posture of the robot in the stored environmental map.An optional implementation for comparing the constructed temporary map with the stored environment map to determine the robot's pose in the stored environment map is as follows: Based on a matching algorithm, the constructed temporary map is traversed over each pose on the stored environment map. For example, if the grid size is 5 cm, a 5 cm step size can be selected to cover all possible poses in the stored environment map. The angle step size is then set to 5 degrees to include the orientation parameters of all poses. When a grid representing an obstacle on the temporary map matches a grid representing an obstacle on the stored environment map, a score is added, and the pose with the highest score is determined as the globally optimal pose. The matching rate of the globally optimal pose is then calculated. When the matching rate of the globally optimal pose exceeds a preset matching rate threshold, the globally optimal pose is determined as the pose information of the autonomous mobile device 10. The pose information of the autonomous mobile device 10 includes its location information and orientation information. After the autonomous mobile device 10 determines its current location, it can match the location with multiple predefined collection locations to determine whether the autonomous mobile device 10 has moved to the predefined collection location. If the location matches the predefined collection locations, it is determined that the autonomous mobile device 10 has moved to the predefined collection location P. Furthermore, the autonomous mobile device 10 can collect the pointer instrument image at the predefined collection location P. In this embodiment, the autonomous mobile device 10 and the server device 20 are communicatively connected. The connection between the autonomous mobile device 10 and the server device 20 can be wireless or wired. Alternatively, the autonomous mobile device 10 and the server device 20 can be communicatively connected via the Internet. Alternatively, the autonomous mobile device 10 and the server device 20 can be communicatively connected to the terminal device 10b via a mobile network. Alternatively, the autonomous mobile device 10 and the server device 20 can be communicatively connected via Bluetooth, Wireless Fidelity (WiFi), or infrared. Based on the communication link between the autonomous mobile device 10 and the server device 20, the autonomous mobile device 10 can provide the captured pointer instrument image to the server device 20, which then reads the pointer instrument based on the pointer instrument image. The server device 20 can be a single server device, a cloud-based server array, or a virtual machine (VM) running within the cloud-based server array. OAlternatively, the server device 20 may also refer to other computing devices with corresponding service capabilities, such as computers or other terminal devices (running service programs). The following exemplifies a method by which the server device 20 reads a pointer instrument based on a pointer instrument image. In various embodiments of the present disclosure, the pointer instrument image includes an image of the dial of the pointer instrument. FIG1 b is a flow chart illustrating a pointer instrument reading method provided in an embodiment of the present disclosure. As shown in FIG1 b , the pointer instrument reading method primarily includes:

[0002] 101. Perform scale number recognition on a pointer instrument image including a dial image of the pointer instrument to determine a scale number recognition result of the pointer instrument.

[0003] 102. Perform pointer detection on the pointer instrument image to determine image position information of a detection frame of the pointer of the pointer instrument.

[0004] 103. Using the image position information of the pointer detection frame as prompt information, perform image segmentation on the pointer instrument image to obtain a target pointer segmentation result.

[0005] 104. Determine target image position information for scale segmentation prompts based on the target pointer segmentation result and the scale number recognition result.

[0006] 105. Using the target image position information as prompt information, perform image segmentation on the pointer instrument image to obtain a target scale segmentation result.

[0007] 106. Determine the reading of the pointer instrument based on the scale numeral recognition result, the target pointer segmentation result, and the target scale segmentation result. For a pointer instrument, the components generally include scale numerals, scales, and a pointer. During manual reading, the reading of the pointer instrument can be determined based on the target scale and the scale numerals that the pointer points to. To determine the reading of the pointer instrument, the scale numerals must be identified from the pointer instrument image. Based on this, in conjunction with step 101 in FIG. 1a and FIG. 1b, scale numeral recognition can be performed on the pointer instrument image to determine the scale numeral recognition result for the pointer instrument. The scale numeral recognition result may include the scale numerals of the pointer instrument, such as the numbers 0-10° on the pointer instrument image in FIG. 1a. Of course, the scale numeral recognition result may also include image position information of the scale numerals, i.e., the position information of the scale numerals on the pointer instrument image. This can be determined using the pixel coordinates corresponding to the scale numerals on the pointer instrument image.

[0008] In the embodiments of the present disclosure, the specific implementation method for recognizing scale digits on a pointer instrument image is not limited. In some embodiments, a digit recognition model can be used to recognize scale digits on a pointer instrument image to determine the scale digit recognition result of the pointer instrument. In the embodiments of the present disclosure, the specific implementation form of the digit recognition model is not limited. Optionally, the digit recognition model can be a neural network model, such as a convolutional neural network (CNN) model, a recurrent neural network (RNN), a deep neural network (DNN), or a feedforward neural network (FNN). Because neural network models for recognizing scale digits require pre-training, a large number of sample images with labeled digits are required. This not only requires a large amount of sample labeling work for model training, but also consumes a high time cost. To reduce model training costs, in some other embodiments, optical character recognition (OCR) technology can be used to perform character recognition on the pointer instrument image to determine the text characters and their image position information. OCR technology detects characters on the pointer instrument image, determines the character shapes by detecting dark and light patterns, and then uses character recognition methods to translate the character shapes into computer text. Specifically, the characters on the pointer instrument image are optically converted into a black and white dot matrix image file, and recognition software is used to convert the text in the image file into text format. Specifically, as shown in step 1 in Figure 1c, OCR technology can be used to perform character detection on the pointer instrument image to determine the image position information of the characters contained in the pointer instrument image. Furthermore, OCR technology can be used to perform character recognition on the characters contained in the pointer instrument image to determine the characters contained in the pointer instrument image, that is, to determine the specific character content contained in the pointer instrument image. Since the dial of a pointer instrument may contain other characters besides scale numbers, such as, but not limited to, the model number, manufacturer, and / or the unit of measurement of the pointer instrument reading, it is necessary to filter out the scale numbers of the pointer instrument from the characters contained in the recognized pointer instrument image.The inventors of the present disclosure have discovered that the positional distribution characteristics of the scale numerals on a pointer instrument's dial, namely, their positional distribution characteristics, follow a certain pattern, generally showing an arc-shaped distribution on the dial. Furthermore, the scale numerals on a pointer instrument exhibit certain logical relationships, such as generally forming an arithmetic progression. Based on this, as shown in step 1 of FIG1c , scale numerals can be selected from the characters in the pointer instrument image based on the characters and their image position information. Specifically, the positional distribution characteristics of the characters can be determined based on their image position information; the logical relationships between the characters can then be determined based on the characters in the pointer instrument image. Subsequently, scale numerals can be selected from the characters in the pointer instrument image based on the character positional distribution characteristics and the logical relationships between the characters. Specifically, characters in the pointer instrument image whose positional distribution characteristics show that they are distributed on the same arc can be selected; and among the characters distributed on the same arc, characters whose logical relationships form an arithmetic progression are selected as the scale numerals for the pointer instrument. After determining the scale numerals of a pointer instrument, image position information of the scale numerals of the pointer instrument can also be obtained from the image position information of the characters contained in the pointer instrument image. The scale numeral recognition results may include the scale numerals of the pointer instrument and their image position information. The image position information of the scale numerals specifically refers to the position information of the scale numerals on the pointer instrument image, generally represented by the pixel coordinates of the scale numerals in the pointer instrument image. The implementation of scale numeral recognition for a pointer instrument image shown in the above embodiment is merely illustrative and not limiting. To determine the reading of a pointer instrument, in addition to recognizing the scale numerals, it is also necessary to obtain information about the pointer and scale of the pointer instrument. Therefore, in the disclosed embodiment, in conjunction with step 102 in Figures 1a and 1b and step 2 in Figure 1b, pointer detection can also be performed on the pointer instrument image to determine image position information of a detection frame for the pointer of the pointer instrument. Optionally, the pointer instrument image can be input into a pointer detection model; within the pointer detection model, pointer detection is performed on the pointer instrument image to obtain image position information of a detection frame for marking the pointer in the pointer instrument image. In practical applications, when performing pointer detection on an image, a rectangular detection frame is typically used to annotate the pointer image contained in the image. The image position information of the rectangular detection frame specifically refers to the image position of the rectangular detection frame on the pointer instrument image, reflecting the spatial distribution of the pointer within the pointer instrument image. The image position information of the detection frame includes information such as the center position and size of the detection frame. Optionally, the center position of the detection frame can be represented by the center coordinates of the rectangular detection frame, and the size of the detection frame can be represented by the width and height of the rectangular detection frame.Accordingly, the image position information of the detection frame can be expressed as (x, y, w, h). (x, y) represents the center coordinates of the rectangular detection frame, i.e., the pixel coordinates of the center of the rectangular detection frame in the pointer instrument image, and w and h represent the width and height of the rectangular detection frame, respectively. Alternatively, the image position information of the rectangular detection frame may include vertex coordinates of the rectangular detection frame. The center coordinates and vertex coordinates of the rectangular detection frame are both pixel coordinates in the pointer instrument image. A pointer instrument image frame may contain one or more pointers. "Multiple" means two or more. Each pointer may correspond to a detection frame. In the disclosed embodiments, before using the pointer detection model for pointer detection, the pointer detection model must be trained. The pointer detection model architecture can include, but is not limited to, a Single Shot Detector (SSD) model, a YOLO (You Only Look Once) model series, a CenterNet model, a Spatial Pyramid Pooling (SPPNet) model, or a Feature Pyramid Network (FPN) model. The battery anomaly recognition model can include, for example, a neural network model. Furthermore, in conjunction with step 103 in Figures 1a and 1b and step 2 in Figure 1c, the image position information of the pointer detection frame can be used as a prompt to perform image segmentation on the pointer instrument image, thereby obtaining the target pointer segmentation result. The image position information of the pointer detection frame provides prompt information for image segmentation. Image segmentation of the pointer instrument image based on this prompt information can complete pointer segmentation in a zero-shot learning environment. This means that pointer segmentation can be achieved for the image of the pointer instrument without having to learn the pointer instrument corresponding to the pointer instrument image. Since learning is not required for the pointer instrument to be read, pointer segmentation can be performed on images of unknown pointer instruments, improving the versatility and universality of pointer segmentation. Specifically, as shown in the image segmentation process corresponding to FIG1C , the image position information of the pointer detection frame can be used as prompt information (Prompt), and the pointer instrument image can be input into the image segmentation model as the image to be segmented. In the image segmentation model, the pointer instrument image and the image position information of the pointer detection frame are respectively encoded to obtain an image feature vector of the pointer instrument image and a prompt feature vector of the pointer detection frame image position information. Furthermore, a mask mapping is performed on the image feature vector of the pointer instrument image and the prompt feature vector of the pointer detection frame image position information to obtain a mask map of the pointer, as shown in FIG2(c). In this embodiment, the image position information of the pointer detection frame is used as prompt information (Prompt), and the image segmentation model is used to perform pointer segmentation on the pointer instrument image. This allows for zero-learning operation to complete pointer segmentation. This means that the image segmentation model can be universal, eliminating the need for training specific to pointer instrument images. This eliminates the need for pointer annotation for images of various types of pointer instruments, reducing the pointer annotation cost of the image segmentation model. Furthermore, because the image segmentation model does not need to be trained specifically for the type of pointer instrument to be read, it can perform pointer segmentation on images of unknown pointer instruments, improving the versatility and universality of pointer segmentation. In the embodiments of the present disclosure, the specific implementation form of the image segmentation model is not limited. The image segmentation model can be a neural network model. In some embodiments, when the number of parameters in the image segmentation model is large, for example, millions, hundreds of millions, billions, or even more, the image segmentation model can also be referred to as a large model. In the embodiments of the present disclosure, a large model is defined as a neural network model whose number of model parameters falls within a preset parameter range.The preset parameter range corresponds to a large number of model parameters, potentially reaching millions, hundreds of millions, billions, or even more. The specific values ​​are determined by standards in the field of artificial intelligence. A large model can be trained through two phases: pre-training and fine-tuning. In the pre-training phase, the model is trained on large-scale, general-domain images to learn the basic structure of language and common sense. Then, in the fine-tuning phase, the model is further trained on smaller, more specific domain datasets. Fine-tuning allows the model to better understand and generate the language of this specific domain, thereby better completing specific tasks. In embodiments of the present disclosure, the image segmentation model can utilize a pre-trained large model, eliminating the need for model learning and training specific to a specific type of pointer instrument. Specifically, the image segmentation model can utilize a general-domain image segmentation model, eliminating the need for pointer annotation for images of various types of pointer instruments, thereby reducing the cost of pointer annotation for the image segmentation model. In some embodiments, the image segmentation model can be a Segment-Anything Model (SAM). The SAM model can perform pointer segmentation with zero-shot learning. As shown in Figure 1c, an image segmentation model (such as a SAM model) may include an image encoder, a prompt word encoder, and a mask decoder. The image encoder encodes the pointer instrument image to obtain an image feature vector. The prompt word encoder is configured to encode the image position information of the pointer detection frame to obtain a prompt feature vector. The mask decoder is configured to map the image feature vector of the pointer instrument image and the prompt feature vector of the pointer detection frame image position information to a mask, perform pointer mask prediction, and obtain a pointer mask map. Furthermore, pointer segmentation may be performed on the pointer instrument image based on the pointer mask map to obtain a pointer segmentation result as shown in Figure 2(d). The pointer segmentation result may include pointer image position information (such as the pixel coordinates of the pointer in the pointer instrument image) and / or a pointer image. The pointer image is shown in Figure 2(d). The pointer segmentation result may be one or more. "Multiple" refers to two or more. For some image segmentation models (such as the SAM model), the pointer segmentation results include multiple pointer segmentation results of different granularities. Pointer segmentation results of different granularities may correspond to different pointer regions.For example, as shown in FIG3 , the pointer segmentation method shown in the aforementioned embodiment is used to perform pointer segmentation on the pointer instrument image shown in FIG3 (a), and three pointer segmentation results of different granularities, namely (1), (2), and (3), are obtained as shown in FIG3 . Different pointer segmentation results contain different pointer areas or granularities. Based on this, a target pointer segmentation result can also be determined from the pointer segmentation results. In some embodiments, there is one pointer segmentation result, and the pointer segmentation result can be determined as the target pointer segmentation result. In other embodiments, the pointer segmentation result is a plurality of pointer segmentation results of different granularities, and the target pointer segmentation result can be determined from the plurality of pointer segmentation results (i.e., “screening target pointer segmentation results” in FIG1C ). Specifically, the pointer segmentation result may include: image position information of the pointer, i.e., pixel coordinates of the pointer. Accordingly, the lengths of the pointers segmented from the multiple pointer segmentation results can be determined based on the image position information of the pointers contained in each of the multiple pointer segmentation results. Furthermore, the pointer segmentation result whose segmented pointer length meets a set length condition can be selected from the multiple pointer segmentation results as the target pointer segmentation result. The set length condition can be a maximum length. Accordingly, the pointer segmentation result whose segmented pointer length is the largest can be selected from the multiple pointer segmentation results as the target pointer segmentation result. Alternatively, the set length condition can be a pointer length within a set length range. The set length range is pre-set based on the actual pointer length of a pointer instrument. Accordingly, the pointer segmentation result whose segmented pointer length is within the set length range can be selected from the multiple pointer segmentation results as the target pointer segmentation result. To determine the reading of a pointer instrument, in addition to identifying the scale numerals and pointer in the pointer instrument image, it is also necessary to obtain information about the pointer instrument's scale. Therefore, it is also necessary to segment the scale from the pointer instrument image. Using a scale segmentation model to segment the scales of a pointer instrument image requires pre-labeling the scales of a specific type of pointer instrument and using these specific pointer instrument images to perform large-scale model training for the scale segmentation model. This scale segmentation approach is associated with high scale labeling costs for training the scale segmentation model and is limited to segmenting specific types of pointer instruments (i.e., those for which the scale segmentation model has been trained). It cannot segment the scales of untrained pointer instruments. To address the aforementioned technical issues, the inventors of the present disclosure discovered that the relative positional relationships between the scales, scale numerals, and pointers of pointer instruments exhibit certain regularities.Based on this, in conjunction with step 104 in Figures 1a and 1b and step 3 in Figure 1c, the target image position information set as a scale segmentation prompt can be determined based on the scale numeral recognition results and target pointer segmentation results obtained in the aforementioned embodiment, i.e., the scale segmentation prompt information determined in Figure 1c. The target pointer segmentation result may include: the image position information of the pointer; the scale numeral recognition result includes: the image position information of the scale numerals. Accordingly, based on the image position information of the pointer and the image position information of the scale numerals obtained in the aforementioned embodiment, multiple pixel coordinates can be determined from the pointer instrument image as the target image position information. Multiple refers to two or more, and the specific number can be flexibly set based on actual needs. The inventors of the present disclosure have discovered that the relative positional relationship between the scale, scale numerals, and pointer of a pointer instrument exhibits a certain regularity. Specifically, as shown in Figures 2 and 3, the scale of a pointer instrument is located between the arc formed by the rotating pointer (defined as the first arc) and the arc formed by the scale numerals (defined as the second arc). Based on this, as shown in Figure 4 , the image position information of the first circular arc formed by the pointer's rotation can be determined based on the image position information of the pointer; and the image position information of the second circular arc formed by the scale numerals can be determined based on the image position information of the scale numerals. Subsequently, based on the image position information of the first and second circular arcs, a target image region located between the first and second circular arcs can be determined from the pointer instrument image. Furthermore, multiple pixel coordinates can be selected from the pixel coordinates of the target image region as the target image position information. For example, in some embodiments, a set number of pixel coordinates can be randomly selected from the pixel coordinates of the target image region as the target image position information; the set number is an integer greater than or equal to 2. Alternatively, pixel coordinates located on a third circular arc having a radius of the target radius r can be determined from the pixel coordinates of the target image region; and multiple pixel coordinates can be randomly selected from the pixel coordinates of the third circular arc as the target image position information. In some embodiments, as shown in FIG4 , if the scale numerals are located outside the scale and the pointer endpoint is located inside the scale, the target radius r is greater than the radius rl of the first arc and less than the radius r2 of the second arc. In other embodiments, as shown in FIG2 and FIG3 , if the scale numerals are located inside the scale and the pointer endpoint is located outside the scale or overlaps with the scale arc, the target radius r is the radius r2 of the second arc and less than the radius rl of the first arc. The target radius r can be any value between the radius rl of the first arc and the radius r2 of the second arc.Alternatively, multiple pixel coordinates can be selected from the pixel coordinates located on the third arc according to a predetermined center angle interval as the target image position information. The above-described embodiments of determining the target image position information for scale segmentation prompts are provided for illustrative purposes only and are not intended to be limiting. Furthermore, in conjunction with steps 105 in Figures 1a and 1b and step 3 in Figure 1c, the target image position information can be used as a prompt to perform image segmentation on the pointer instrument image to obtain a target scale segmentation result. The target image position information provides prompt information for scale segmentation. Scale segmentation of the pointer instrument image based on this prompt information can be completed using zero-shot learning. This means that scale segmentation of the pointer instrument image can be achieved without learning the pointer instrument corresponding to the pointer instrument image. Since learning is not required for the pointer instrument to be read, scale segmentation can be performed on images of unknown pointer instruments, improving the versatility and universality of scale segmentation. Specifically, as shown in the image segmentation process in Figure 1c, the target image location information can be used as a prompt, and the pointer instrument image can be used as the image to be segmented. The image segmentation model then encodes the pointer instrument image and the target image location information to obtain an image feature vector for the pointer instrument image and a prompt feature vector for the target image location information. Furthermore, mask mapping is performed on the image feature vector for the pointer instrument image and the prompt feature vector for the target image location information to obtain a scale mask. In this embodiment, the image segmentation model uses the target image location information as a prompt to perform scale segmentation on the pointer instrument image. This allows for zero-learning implementation. The image segmentation model can be universal, eliminating the need for training specific to pointer instrument images. This eliminates the need for scale annotation for various pointer instrument images, reducing the scale annotation cost of the image segmentation model. Furthermore, since the image segmentation model does not need to be trained specifically for the type of pointer instrument to be read, it can perform scale segmentation on images of unknown pointer instruments, thereby improving the versatility and universality of scale segmentation. The specific implementation of the image segmentation model can be found in the relevant content of the aforementioned embodiments and will not be further elaborated here. After determining the scale mask image, the scale segmentation can be performed on the pointer instrument image based on the scale mask image to obtain a scale segmentation result. The scale segmentation result may include: image position information of the scale (e.g., pixel coordinates of the scale in the pointer instrument image) and / or an image of the scale.The image of the pointer is shown in Figure 5(b). The scale segmentation results may be one or more. "More" means two or more. For the SAM model, the scale segmentation results include multiple scale segmentation results of different granularities. Scale segmentation results of different granularities may correspond to different scale areas. Different scale segmentation results may contain different scale areas or granularities. Based on this, a target scale segmentation result can be determined from the scale segmentation results. In some embodiments, if there is one scale segmentation result, this scale segmentation result can be determined as the target scale segmentation result. In other embodiments, if there are multiple scale segmentation results of different granularities, the target scale segmentation result can be determined from the multiple scale segmentation results (corresponding to the screening of the target scale segmentation result in Figure 1c). Specifically, the scale segmentation results may include: scale image position information, i.e., scale pixel coordinates and / or scale images. The inventors of the present disclosure have discovered that the scales of pointer instruments are generally distributed in a circular arc pattern. Based on this, arc fitting can be performed on the scale distribution shapes corresponding to each of the multiple scale segmentation results based on the image position information of the scales contained in each of the multiple scale segmentation results to obtain the arc fit of the scale distribution shapes corresponding to each of the multiple scale segmentation results. The arc fit can be represented by the square of the correlation coefficient R (R-Square) obtained by performing arc fitting on the scale distribution shape corresponding to the scale segmentation result X. R-Square is a value that measures the degree of correlation between variables, also known as goodness of fit or coefficient of determination. It is set to represent the percentage of variation in the dependent variable that the fitted arc can explain. The closer R-Squard is to 1, the better the fit. After obtaining the arc fit of the scale distribution shapes corresponding to each of the multiple scale segmentation results, the scale segmentation result whose arc fit of the scale distribution shape meets a set arc fit condition can be selected from the multiple scale segmentation results as the target scale segmentation result. The set arc fit condition can be that the larger the arc fit, the better the arc fit. Accordingly, from the multiple scale segmentation results, the scale segmentation result with the highest arc fit of the scale distribution shape can be selected as the target scale segmentation result. Alternatively, the set arc fit condition can be that the arc fit is greater than or equal to a set arc fit threshold. Accordingly, from the multiple scale segmentation results, the scale segmentation result with the arc fit of the scale distribution shape greater than or equal to the set arc fit threshold can be selected as the target scale segmentation result.If there are multiple scale segmentation results whose arc fit of the scale distribution shape is greater than or equal to a set arc fit threshold, the scale segmentation result with the largest arc fit among the scale segmentation results whose arc fit is greater than or equal to the set arc fit threshold may be selected as the target scale segmentation result. In other embodiments, when determining the target scale segmentation result from multiple scale segmentation results, blank areas corresponding to the multiple scale segmentation results may be identified from the images of the scales respectively included in the multiple scale segmentation results, and the areas of the blank areas corresponding to the multiple scale segmentation results may be calculated. Based on the areas of the blank areas corresponding to the multiple scale segmentation results, the scale segmentation result whose blank area satisfies a set area condition may be selected from the multiple scale segmentation results as the target scale segmentation result. The set area condition may be the largest length. Accordingly, the scale segmentation result with the largest blank area among the multiple scale segmentation results may be selected as the target scale segmentation result. Alternatively, the set area condition may be that the blank area falls within a set area range. The set area range can be flexibly configured based on actual needs. Accordingly, from multiple scale segmentation results, the scale segmentation result with a blank area within the set area range can be selected as the target scale segmentation result. If multiple scale segmentation results have blank areas within the set area range, the scale segmentation result with the largest blank area can be selected as the target scale segmentation result. In yet other embodiments, the aforementioned methods of selecting a target scale segmentation result from multiple scale segmentation results based on the arc fit of the scale distribution shape corresponding to the scale segmentation result and selecting a target scale segmentation result from multiple scale segmentation results based on the area of ​​the blank area corresponding to the scale segmentation result can be combined. Specifically, from multiple scale segmentation results, the scale segmentation result with a scale distribution shape that satisfies the set arc fit condition and a blank area that satisfies the set area condition can be selected as the target scale segmentation result. The above-described embodiments of determining a target scale segmentation result for a pointer instrument image are merely illustrative and not limiting. Since the pointer reading is determined by the scale numerals, the scale of the pointer instrument, and the pointer position of the pointer instrument, after determining the scale numeral recognition result, the target scale segmentation result, and the target pointer segmentation result, the pointer instrument reading can be determined based on the scale numeral recognition result, the target pointer segmentation result, and the target scale segmentation result.Specifically, since the scale number recognition result includes the scale numbers of the pointer instrument, the range of the pointer instrument can be determined based on the scale number recognition result (specifically, the scale numbers). Furthermore, in conjunction with step 106 in Figures 1a and 1b, the scale division value of the pointer instrument can be determined based on the target scale segmentation result and the range of the pointer instrument. The scale division value is the minimum scale value of the pointer instrument, the minimum value that can be read on the pointer instrument, and the value of the smallest grid between two adjacent scale divisions on the pointer instrument. This process is not shown in Figure 1c. Specifically, the number of scale divisions included in the target scale segmentation result can be determined based on the scale image in the target scale segmentation result. Subsequently, the scale division value of the pointer instrument can be determined based on the range of the pointer instrument and the number of scale divisions included in the target scale segmentation result. The scale division value of the pointer instrument is the quotient of the range of the pointer instrument divided by the number of scale divisions included in the target scale segmentation result. After determining the graduation values ​​of the pointer instrument, the target scale at which the pointer is pointing can be determined based on the target scale segmentation results and the target pointer recognition results. Specifically, based on the image position information of the scale in the target scale segmentation results and the image position information of the pointer in the target pointer recognition results, the scale whose image position information overlaps with the image position information of the pointer can be determined from the pointer instrument scale. This is the target scale at which the pointer is pointing, thereby determining the specific scale at which the pointer is pointing. Subsequently, based on the graduation values ​​of the pointer instrument and the target scale, the reading corresponding to the target scale can be determined, which is the reading of the pointer instrument. Specifically, the graduation value of the pointer instrument can be multiplied by the sequence of the target scale (i.e., the number of the target scale) to obtain the reading corresponding to the target scale, which is the reading of the pointer instrument. In the disclosed embodiment, scale digit recognition is performed on a pointer instrument image to obtain a scale digit recognition result; pointer segmentation is performed on the pointer instrument image using the image position information of the detection frame of the pointer of the pointer instrument as prompt information; and scale segmentation is performed on the pointer instrument image using the target image position information determined based on the pointer segmentation result and the scale digit recognition result as prompt information. This allows for the segmentation of the pointer and scale without any learning required, and can also achieve semantic segmentation of the pointer instrument image without the need to learn the pointer instrument corresponding to the pointer instrument image. Since there is no need to learn specifically for the pointer instrument to be read, pointer segmentation can be performed on images of pointer instruments of unknown types, thereby improving the versatility and universality of semantic segmentation of pointer instrument images, and thereby improving the versatility and universality of pointer instrument readings.On the other hand, in the disclosed embodiments, only the sample images used by the pointer detection model for pointer detection require pointer annotation. The image segmentation model used for pointer and scale segmentation does not require annotation. This eliminates the need for semantic segmentation sample annotation, reducing sample annotation workload and cost. Furthermore, because the disclosed embodiments perform scale digit recognition on pointer instrument images to obtain a scale digit recognition result, which reflects the range of the pointer instrument, the scale digit recognition result can be used to determine the range of the pointer instrument. When subsequently reading the pointer instrument, the pointer instrument can be read based on the range determined from the scale digit recognition result, the scale segmentation result, and the pointer segmentation result. Therefore, the pointer instrument reading method provided by the disclosed embodiments can also be used to read pointer instruments with unknown ranges, further improving the versatility and universality of pointer instrument reading. Because the target image position information used as a scale segmentation hint is determined based on the scale digit recognition result and the pointer segmentation result, it has a certain degree of randomness and uncertainty. Consequently, the accuracy of the scale segmentation result derived from the target image position information also has a certain degree of uncertainty. Therefore, to improve the accuracy of pointer instrument readings, the target image position information set as the scale segmentation hint can also be optimized. Specifically, as shown in step 4 of FIG1c , a reliability score for the target scale segmentation result can be determined based on the target scale segmentation result. In the disclosed embodiments, the specific implementation for determining the reliability score for the target scale segmentation result based on the target scale segmentation result is not limited. Several implementations are described below as examples. Implementation 1: A scale uniformity score can be determined based on the target scale segmentation result; and based on the scale uniformity score, a reliability score for the target scale segmentation result can be determined. The reliability score for the target scale segmentation result is positively correlated with the scale uniformity score. That is, the higher the scale uniformity score (the more uniform the scale distribution), the higher the reliability score for the target scale segmentation result. In some embodiments, the scale uniformity score can be used as the reliability score for the target scale segmentation result. Specifically, the pixel spacing between adjacent scales can be determined based on the scale image position information included in the target scale segmentation result. The scale uniformity score can then be determined based on the difference in pixel spacing between two adjacent scales. The difference in pixel spacing between two adjacent scales can be represented by the mean squared difference of the pixel spacing between two adjacent scales. For example, the difference in pixel spacing between two adjacent scales can be expressed as:

[0009] 1 N-1 - > - =^Z(j(X,a - X] +(J-D)2 - J(X~ - X,J +(D2 -D)2)2⑵. In formula (2), N represents the total number of scales. i represents the i-th scale, i=O,l, (Nl)o (X, YD represents the image position information of the i-th scale; (K+1, Y 1+ i) A graph representing the (i+l)th scale The uniformity score of the scale can be determined based on the difference in pixel spacing between black lines in the binary image of the scale. The uniformity score of the scale is inversely correlated with the difference in pixel spacing between black lines. That is, the smaller the difference in pixel spacing between two adjacent black lines, the higher the uniformity score of the scale. The difference in pixel spacing between two adjacent scale lines can be represented by the mean squared difference of the pixel spacing between two adjacent black lines. Optionally, a mathematical relationship model between the uniformity score of the scale and the difference in pixel spacing between two adjacent black lines can be pre-set. In this mathematical relationship model, the uniformity score of the scale is the dependent variable, the difference in pixel spacing between two adjacent black lines is the independent variable, and the dependent and independent variables are inversely correlated. In yet other embodiments, the number of scale marks between two adjacent scale numerals can be determined based on the image position information of the scale and the scale numeral recognition results. Specifically, the number of scale marks between two adjacent scale numerals can be determined based on the image position information of the scale and the image position information of the scale numerals. Furthermore, a uniformity score for the scale line can be determined based on the difference in the number of scale marks between every two adjacent scale numerals. The uniformity score of the scale line is inversely correlated with the difference in the number of scale marks between every two adjacent scale numerals. That is, the smaller the difference in the number of scale marks between every two adjacent scale numerals, the higher the uniformity score. Alternatively, the difference in the number of scale marks between every two adjacent scale numerals can be represented by the mean squared error of the number of scale marks between every two adjacent scale numerals. For example, the difference in the number of scale marks between every two adjacent scale numerals can be expressed as:

[0010] ( 3) . In formula (3), M represents the total number of scale numbers. i represents the i-th scale number, i=O,l, (MD o

[0011] " represents the number of scales between the (i+1)th scale number and the i-th scale number. "a represents the number of scales between the (i+2)th scale number and the (i+1)th scale number. Correspondingly, AD represents the mean square error of the number of scales between every two adjacent scale numbers. Optionally, a mathematical relationship model between the uniformity score of the scale and the difference in the number of scales between every two adjacent scale numbers can be preset. In the mathematical relationship model, the uniformity score of the scale is the dependent variable, the difference in the number of scales between every two adjacent scale numbers is the independent variable, and the dependent variable is inversely correlated with the independent variable. The implementation method of determining the uniformity score of the scale shown in the above embodiment is only an example description and does not constitute a limitation. In some embodiments, the reliability score of the target scale segmentation result can be determined based on the uniformity score of the scale. The reliability score of the target scale segmentation result is positively correlated with the scale uniformity score. That is, the higher the scale uniformity score (the more uniform the scale distribution), the higher the reliability score of the target scale segmentation result. In some embodiments, the scale uniformity score can be used as the reliability score of the target scale segmentation result. In other embodiments, a mathematical relationship model between the reliability score of the target scale segmentation result and the scale uniformity score can be pre-set. In this mathematical relationship model, the scale uniformity score is the independent variable, and the reliability score of the target scale segmentation result is the dependent variable, with the dependent variable positively correlated with the independent variable. In this embodiment, the scale uniformity score can be substituted into the mathematical relationship model to obtain the reliability score of the target scale segmentation result. Embodiment 2: Determine the arc fit of the scale distribution shape based on the target scale segmentation result. Specifically, arc fitting can be performed on the scale distribution shape based on the scale position distribution information included in the target scale segmentation result to obtain the arc fit of the scale distribution shape. For details on the implementation of arc fit, please refer to the relevant content of the aforementioned embodiments and will not be elaborated upon here. Furthermore, the reliability score of the target scale segmentation result can be determined based on the arc fit of the scale distribution shape. The reliability score of the target scale segmentation result is positively correlated with the arc fit of the scale distribution shape. That is, the greater the arc fit of the scale distribution shape, the higher the reliability score of the target scale segmentation result. In some embodiments, a mathematical relationship model between the reliability score of the target scale segmentation result and the arc fit of the scale distribution shape can be pre-set. In this mathematical relationship model, the arc fit of the scale distribution shape is the independent variable, the reliability score of the target scale segmentation result is the dependent variable, and the dependent variable is positively correlated with the independent variable.In this embodiment, the arc fit of the scale distribution shape can be incorporated into a mathematical relationship model to obtain the reliability score of the target scale segmentation result. Embodiment 3: The arc fit of the scale distribution shape and the scale uniformity score can be combined to determine the reliability score of the target scale segmentation result. Specifically, the reliability score of the target scale segmentation result can be determined based on the scale uniformity score and the arc fit of the scale distribution shape. In some embodiments, a weighted sum of the scale uniformity score and the arc fit of the scale distribution shape can be performed to obtain the reliability score of the target scale segmentation result. Alternatively, a mathematical relationship model can be pre-set between the scale uniformity score, the arc fit of the scale distribution shape, and the reliability score of the target scale segmentation result. In this mathematical relationship model, the scale uniformity score and the arc fit of the scale distribution shape are independent variables, and the reliability score of the target scale segmentation result is the dependent variable, with the independent and dependent variables being positively correlated. Based on this mathematical relationship model, the scale uniformity score and arc fit of the scale distribution shape obtained using the above-described embodiments 1 and 2 can be substituted into the mathematical relationship model for solution to obtain a reliability score for the target scale segmentation result. The above-described embodiments for determining the reliability score of the target scale segmentation result are merely illustrative and not limiting. In the present disclosure, the target image position information set as a scale segmentation prompt can be optimized based on the reliability score of the target scale segmentation result. Specifically, as shown in step 4 of FIG1c , the target image position information can be iteratively adjusted using an optimization method to obtain the adjusted image position information. For example, the target image position information can be iteratively adjusted using a gradient descent method to obtain the adjusted image position information. After each adjustment of the target image position information, the pointer instrument image can be segmented using the adjusted image position information as the prompt information to obtain the target scale segmentation result corresponding to each adjusted image position information. For the specific implementation of performing image segmentation on the pointer instrument image using the image position information after each adjustment as prompt information, please refer to the aforementioned description of performing image segmentation on the pointer instrument image using the target image position information as prompt information, and will not be repeated here. Furthermore, based on the target scale segmentation result corresponding to each adjusted image position information, a reliability score for the target scale segmentation result corresponding to each adjusted image position information can be determined. The specific implementation of this step can refer to the aforementioned description of determining the reliability score for the target scale segmentation result based on the target scale segmentation result, and will not be repeated here.Furthermore, based on the reliability score of the target scale segmentation result and the reliability score of the target scale segmentation result corresponding to each adjusted target image position information, a target scale segmentation result whose reliability score meets the set reliability score condition can be determined from the target scale segmentation results and the target scale segmentation results corresponding to each adjusted target image position information. For example, as shown in step 4 of FIG1c , the target scale segmentation result with the highest reliability score can be selected from the target scale segmentation results and the target scale segmentation results corresponding to each adjusted target image position information as the target scale segmentation result whose reliability score meets the set reliability score condition, and so on. Furthermore, the reading of the pointer instrument can be determined based on the target scale segmentation result whose reliability score meets the set reliability score condition, the aforementioned scale numeral recognition result, and the target pointer segmentation result. For the specific implementation of this step, please refer to the aforementioned section regarding determining the reading of the pointer instrument based on the target scale segmentation result, the aforementioned scale numeral recognition result, and the target pointer segmentation result, and will not be further elaborated here. In this embodiment, an optimization method is used to iteratively obtain a scale segmentation result whose reliability score satisfies a preset reliability score condition (e.g., the highest reliability score). The reliability of the scale segmentation result reflects the accuracy of the scale segmentation. Therefore, this embodiment can improve the accuracy of scale segmentation, and thus improve the accuracy of subsequent pointer instrument readings. It is worth noting that the aforementioned method of performing pointer instrument readings based on pointer instrument images by the server device 20 can also be deployed on an autonomous mobile device 10, with the autonomous mobile device 10 autonomously completing the processes of capturing pointer instrument images and performing pointer instrument readings based on the pointer instrument images. This process does not require the involvement of the server device 20. For the specific implementation of pointer instrument readings based on pointer instrument images by the autonomous mobile device 10, please refer to the aforementioned related content regarding the server device 20 performing pointer instrument readings based on pointer instrument images, and will not be further described here. It is also worth noting that the pointer instrument reading method provided in the embodiments of the present disclosure is not limited to reading pointer instruments in data centers and can be adapted for reading pointer instruments in any scenario. For example, it can be used to read pointer instruments in any industrial scenario. In addition, the execution entity of the pointer instrument reading method provided in the embodiments of the present disclosure can be any device with computing capabilities, including but not limited to: terminal devices such as desktop computers, laptops, mobile phones, or IoT devices; it can also be various server devices such as traditional servers, cloud servers, or server clusters.It should be noted that the execution entity of each step of the method provided in the above embodiments can be the same device, or the method can be executed by different devices. For example, the execution entity of steps 101 and 102 can be device A; for another example, the execution entity of step 101 can be device A, and the execution entity of step 102 can be device B, and so on. In addition, some processes described in the above embodiments and figures include multiple operations that appear in a specific order. However, it should be understood that these operations may not be executed in the order in which they appear herein or may be executed in parallel. The sequence numbers of the operations, such as 101 and 102, are merely provided to distinguish different operations and do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. Accordingly, embodiments of the present disclosure further provide a computer-readable storage medium storing computer instructions. When the computer instructions are executed by one or more processors, the one or more processors execute the steps of the above-mentioned pointer instrument reading method. Figure 6 is a schematic structural diagram of an electronic device provided in embodiments of the present disclosure. As shown in Figure 6 , the electronic device may include a memory 60a and a processor 60b. The memory 60a is configured to store a computer program. The processor 60b is coupled to the memory 60a and configured to execute the computer program to perform the steps of the pointer instrument reading method provided in the aforementioned embodiments. The specific implementation of each step can be found in the relevant descriptions of the aforementioned embodiments and will not be repeated here. In some optional embodiments, as shown in Figure 6 , the electronic device may further include optional components such as a communication component 60c, a power component 60d, a display component 60e, and an audio component 60f. In some embodiments, the electronic device may further include an image acquisition device configured to acquire an image of the pointer instrument, etc. Figure 6 only schematically illustrates some components and does not imply that the electronic device must include all components shown in Figure 6 , nor does it imply that the electronic device must include only the components shown in Figure 6 . Furthermore, the components within the dashed boxes in Figure 6 are optional components, not mandatory components, and may depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, laptop computer, mobile phone, or IoT device; it can also be various server devices such as a traditional server, cloud server, or server cluster. The electronic device can also be implemented as an autonomous mobile device. Accordingly, it can also include: a driver component (not shown in FIG. 6 ), etc.The drive components may include drive wheels, drive motors, universal wheels, etc. The basic components and their composition may vary between different autonomous mobile devices; the examples listed in this disclosure are merely some examples. In this disclosure, the memory is configured to store computer programs and may also be configured to store various other data to support operations on the device in which it is located. The processor may execute the computer program stored in the memory to implement the corresponding control logic. The memory may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. In the embodiments of the present disclosure, the processor may be any hardware processing device capable of executing the logic of the aforementioned method. Optionally, the processor may be a central processing unit (CPU), a graphics processing unit (GPU), or a microcontroller unit (MCU); may also be a programmable device such as a field-programmable gate array (FPGA), a programmable array logic device (PAL), a general array logic device (GAL), or a complex programmable logic device (CPLD); or may be an advanced reduced instruction set compute (RISC) processor (ARM) or a system on a chip (SoC), etc., but is not limited thereto.In an embodiment of the present disclosure, a communication component is configured to facilitate wired or wireless communication between the device in which it resides and other devices. The device in which the communication component resides can access a wireless network based on a communication standard, such as Wireless Fidelity (WiFi), 2G or 3G, 4G, 5G, or a combination thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In another exemplary embodiment, the communication component may also be implemented based on near-field communication (NFC) technology, radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, or other technologies. In an embodiment of the present disclosure, the display component may include a liquid crystal display (LCD) and a touch panel (TP). If the display component includes a touch panel, the display component may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensors can not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide action. In embodiments of the present disclosure, a power supply component is configured to provide power to various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device in which the power supply component resides. In embodiments of the present disclosure, an audio component may be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC). When the device in which the audio component resides is in an operating mode, such as call mode, recording mode, or voice recognition mode, the microphone is configured to receive external audio signals. The received audio signals may be further stored in a memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker configured to output audio signals. For example, for devices with voice interaction capabilities, voice interaction with the user can be implemented through the audio component.It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data configured for analysis, storage, and display) involved in this disclosure are all authorized by the user or fully authorized by all parties. The collection, use, and processing of relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or deny. It should also be noted that the descriptions of "first" and "second" herein are used to distinguish different messages, devices, components, etc., and do not imply a sequential order, nor do they limit "first" and "second" to different types. Those skilled in the art will understand that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware. Furthermore, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, compact disc read-only memory (CD-ROM), optical storage, etc.) containing computer-usable program code. The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, such that the instructions, when executed by the processor of the computer or other programmable data processing device, produce a device configured to implement the functions specified in one or more processes in the flowcharts and / or one or more blocks in the block diagrams. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, which implements the functions specified in one or more processes in the flowchart and / or one or more blocks in the block diagram.These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing the computer or other programmable device to execute a series of operational steps to produce a computer-implemented process. The instructions executed on the computer or other programmable device provide steps configured to implement the functions specified in one or more flow charts and / or one or more blocks in a block diagram. In a typical configuration, a computing device includes one or more processors (CPU, etc.), input / output interfaces, network interfaces, and memory. Memory may include non-permanent storage in a computer-readable medium, random-access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash memory. Memory is an example of a computer-readable medium. Computer storage media is readable storage media, also referred to as readable media. Readable storage media, including both permanent and non-permanent, removable and non-removable media, can be implemented using any method or technology to store information. Information can be computer-readable instructions, data structures, program components, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic tape cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be configured to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves. It should also be noted that the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, product or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, product or apparatus.In the absence of further limitations, elements defined by the phrase "comprising a..." do not preclude the presence of additional identical elements in the process, method, product, or device comprising the aforementioned elements. The above content is merely an embodiment of the present disclosure and is not intended to limit the present disclosure. Various modifications and variations are readily apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure are intended to be encompassed by the claims of the present disclosure. Industrial Applicability: The pointer instrument reading method, device, system, and storage medium provided by the embodiments of the present disclosure are configured in the field of image processing technology. The method, device, system, and storage medium utilize scale digit recognition on a pointer instrument image to obtain a scale digit recognition result. The method then uses the image position information of the pointer detection frame of the pointer instrument as prompt information to segment the pointer on the pointer instrument image. Finally, the method uses the target image position information determined based on the pointer segmentation result and the scale digit recognition result as prompt information to segment the scale on the pointer instrument image. This method can achieve zero-learning segmentation of the pointer and scale, eliminating the need to learn the pointer instrument corresponding to the pointer instrument image. Furthermore, semantic segmentation of the pointer instrument image can be achieved. Since learning for a specific pointer instrument type is not required, pointer segmentation can be performed on images of unknown pointer instruments, improving the versatility of semantic segmentation of pointer instrument images and, in turn, improving the versatility of pointer instrument reading.

Claims

Claims: A pointer instrument reading method, comprising: Performing scale digital recognition on a pointer instrument image including a dial image of a pointer instrument to determine a scale digital recognition result of the pointer instrument; Performing pointer detection on the pointer instrument image to determine image position information of a detection frame of a pointer of the pointer instrument; Using the image position information of the detection frame of the pointer as prompt information, performing image segmentation on the pointer instrument image to obtain a target pointer segmentation result; determining the target image position information set as a scale segmentation prompt according to the target pointer segmentation result and the scale number recognition result; Using the target image position information as prompt information, performing image segmentation on the pointer instrument image to obtain a target scale segmentation result; The reading of the pointer instrument is determined according to the scale digit recognition result, the target pointer segmentation result and the target scale segmentation result. , The method according to claim 1, wherein, The step of performing scale digit recognition on a pointer instrument image of a dial image including a pointer instrument to determine a scale digit recognition result of the pointer instrument includes: performing character recognition on the pointer instrument image using an optical character recognition technology to determine characters included in the pointer instrument image and image position information of the characters; filtering out the scale digits from the characters included in the pointer instrument image according to the characters included in the pointer instrument image and the image position information of the characters; and acquiring image position information of the scale digits from the image position information of the characters; the scale digit recognition result includes: the scale digits and the image position information of the scale digits. , The method according to claim 2, wherein, The step of selecting the scale numbers from the characters included in the pointer instrument image according to the characters included in the pointer instrument image and the image position information of the characters comprises: determining position distribution features of the characters according to the image position information of the characters; determining a logical relationship between the characters according to the characters included in the pointer instrument image; and selecting the scale numbers from the characters included in the pointer instrument image according to the position distribution features of the characters and the logical relationship between the characters. , The method according to claim 3, wherein, The step of selecting the scale numbers from the characters included in the pointer instrument image according to the position distribution characteristics of the characters and the logical relationship between the characters includes: selecting characters whose position distribution characteristics are distributed on the same arc from the characters included in the pointer instrument image; and selecting characters whose logical relationship is an arithmetic progression from the characters distributed on the same arc as the scale numbers. , The method according to claim 1, wherein, The method of performing image segmentation on the pointer instrument image using the image position information of the pointer as prompt information to obtain a target pointer segmentation result includes: using the image position information of the detection frame of the pointer as prompt information and the pointer instrument image as the image to be segmented to input an image segmentation model; encoding the image position information of the pointer instrument image and the detection frame of the pointer in the image segmentation model respectively to obtain an image feature vector of the pointer instrument image and a prompt feature vector of the image position information of the detection frame of the pointer; performing mask mapping on the image feature vector of the pointer instrument image and the prompt feature vector of the image position information of the detection frame of the pointer to obtain a mask map of the pointer; performing image segmentation on the pointer instrument image according to the mask map of the pointer to determine a pointer segmentation result; The target pointer segmentation result is determined from the pointer segmentation results. , The method according to claim 5, wherein, The pointer segmentation result includes: multiple pointer segmentation results of different granularities; the pointer segmentation result includes: image position information of the pointer; determining the target pointer segmentation result from the pointer segmentation results includes: determining the length of the pointer segmented by the multiple pointer segmentation results according to the image position information of the pointer respectively contained in the multiple pointer segmentation results; from the multiple pointer segmentation results, selecting the pointer segmentation result whose length of the segmented pointer meets the set length condition as the target pointer segmentation result. , The method according to claim 1, wherein, The target pointer segmentation result includes: image position information of the pointer; the scale number recognition result includes: image position information of the scale number; the determining the target image position information set as the scale segmentation prompt according to the pointer segmentation result and the scale number recognition result includes: determining the image position information of a first arc formed by the rotation of the pointer according to the image position information of the pointer; determining the image position information of a second arc formed by the scale number according to the image position information of the scale number; determining a target image area between the first arc and the second arc from a pointer instrument image according to the image position information of the first arc and the image position information of the second arc; and selecting a plurality of pixel coordinates from the pixel coordinates of the target image area as the target image position information. , The method according to claim 7, wherein, The selecting the multiple pixel coordinates from the pixel coordinates of the target image area includes: randomly selecting a set number of pixel coordinates from the pixel coordinates of the target image area; the set number is greater than or equal to 2 and is an integer; or, determining the pixel coordinates located on a third arc with a radius of a target radius from the pixel coordinates of the target image area; randomly selecting the multiple pixel coordinates from the pixel coordinates located on the third arc; or, determining the pixel coordinates located on a third arc with a radius of a target radius from the pixel coordinates of the target image area; selecting the multiple pixel coordinates from the pixel coordinates located on the third arc according to a set center angle interval; wherein the target radius is greater than the radius of the first arc and less than the radius of the second arc, or the target radius is greater than the radius of the second arc and less than the radius of the first arc. , The method according to claim 1, wherein, The method of performing image segmentation on the pointer instrument image with the target image position information as prompt information to obtain a scale segmentation result includes: inputting an image segmentation model with the target image position information as prompt information and the pointer instrument image as the image to be segmented; encoding the pointer instrument image and the target image position information in the image segmentation model respectively to obtain an image feature vector of the pointer instrument image and a prompt feature vector of the target image position information; performing mask mapping on the image feature vector of the pointer instrument image and the prompt feature vector of the target image position information to obtain a mask map of the scale of the pointer instrument; performing image segmentation on the pointer instrument image according to the mask map of the scale to determine the scale segmentation result; and determining the target scale segmentation result from the scale segmentation result. , The method according to claim 9, wherein, The scale segmentation result includes: a plurality of scale segmentation results of different granularities; the scale segmentation result includes: image position information of the scale and / or an image of the scale; determining the target scale segmentation result from the scale segmentation results includes: performing arc fitting on the scale distribution shapes corresponding to the plurality of scale segmentation results according to the image position information of the scales respectively included in the plurality of scale segmentation results, so as to obtain arc fitting degrees of the scale distribution shapes respectively corresponding to the plurality of scale segmentation results; selecting, from the plurality of scale segmentation results, a scale segmentation result whose arc fitting degree satisfies a set arc fitting degree condition according to the arc fitting degrees of the scale distribution shapes respectively corresponding to the plurality of scale segmentation results, as the target scale segmentation result; and / or, identifying blank areas corresponding to the plurality of scale segmentation results from the images of the scales respectively included in the plurality of scale segmentation results, and calculating the areas of the blank areas corresponding to the plurality of scale segmentation results; and calculating, according to the areas of the blank areas corresponding to the plurality of scale segmentation results, A scale segmentation result in which the area of ​​the blank region meets a set area condition is selected from the plurality of scale segmentation results as the target scale segmentation result. , The method according to any one of claims 1 to 10, wherein, The determining the reading of the pointer instrument according to the scale digit recognition result, the target pointer segmentation result and the target scale segmentation result comprises: determining the reliability score of the target scale segmentation result according to the target scale segmentation result; iteratively adjusting the target image position information through an optimization method to obtain the image position information after each adjustment; performing image segmentation on the pointer instrument image with the image position information after each adjustment as prompt information to obtain the target scale segmentation result corresponding to the image position information after each adjustment; determining the reliability score of the target scale segmentation result corresponding to the image position information after each adjustment according to the target scale segmentation result corresponding to the image position information after each adjustment; determining the target scale segmentation result whose reliability score meets the set reliability score condition from the target scale segmentation result and the target scale segmentation result corresponding to the target image position information after each adjustment according to the reliability score of the target scale segmentation result and the reliability score of the target scale segmentation result corresponding to the target image position information after each adjustment; determining the reading of the pointer instrument according to the target scale segmentation result meeting the set reliability score condition, the scale digit recognition result and the target pointer segmentation result. , The method according to claim 11, wherein, Determining the reliability score of the target scale segmentation result according to the target scale segmentation result includes: determining the uniformity score of the scale according to the target scale segmentation result; determining the reliability score of the target scale segmentation result according to the uniformity score of the scale; or, determining the arc fitting of the distribution shape of the scale according to the target scale segmentation result; determining the reliability score of the target scale segmentation result according to the arc fitting of the distribution shape of the scale; or, determining the reliability score of the target scale segmentation result according to the uniformity score of the scale and the arc fitting of the distribution shape of the scale; wherein, the reliability score of the target scale segmentation result is positively correlated with the uniformity score of the scale; and the reliability score of the target scale segmentation result is positively correlated with the arc fitting of the distribution shape of the scale. , The method according to claim 12, wherein, The target scale segmentation result includes: image position information of the scale and / or an image of the scale; determining the uniformity score of the scale according to the target scale segmentation result includes: determining the pixel spacing between adjacent scales according to the image position information of the scale; determining the uniformity score of the scale according to the difference in pixel spacing between every two adjacent scales; wherein the uniformity score of the scale is inversely correlated with the difference in pixel spacing between every two adjacent scales; 19 Alternatively, binarization is performed on the image of the scale to obtain a binarized image of the scale; a uniformity score of the scale is determined according to a difference between pixel intervals between black lines in the binarized image of the scale; wherein the uniformity score of the scale is anti-correlated with the difference between pixel intervals between black lines; or, the number of scales between two adjacent scale numbers is determined according to image position information of the scale and the scale number recognition result; a uniformity score of the scale line is determined according to a difference in the number of scales between every two adjacent scale numbers; wherein the uniformity score of the scale is anti-correlated with the difference in the number of scales between every two adjacent scale numbers. , The method according to claim 11, wherein, Determining the reading of the pointer instrument according to the target scale segmentation result that meets the set reliability score condition, the scale number recognition result and the target pointer segmentation result includes: determining the range of the pointer instrument according to the scale number recognition result; determining the graduation value of the pointer instrument according to the target scale segmentation result that meets the set reliability score condition and the range of the pointer instrument; determining the target scale pointed to by the pointer according to the target scale segmentation result that meets the set reliability score condition and the target pointer recognition result; determining the reading corresponding to the target scale according to the graduation value of the pointer instrument and the target scale as the reading of the pointer instrument. , A data center inspection system, comprising: Autonomous mobile devices and server devices; The autonomous mobile device is provided with an image acquisition device; The autonomous mobile device is configured to move in a data center, and during the movement, controls the image acquisition device to acquire images of pointer instruments in the data center to obtain a pointer instrument image including a dial image of the pointer instrument; and provides the pointer instrument image to the server device; the server device is configured to execute the steps in the method according to any one of claims 1 to 14. , an electronic device, comprising: A memory and a processor; wherein the memory is configured to store a computer program; and the processor is coupled to the memory and configured to execute the computer program to execute the steps in the method according to any one of claims 1 to 14. A computer-readable storage medium storing computer instructions, when the computer instructions are executed by one or more processors, causes the one or more processors to execute the steps in any one of the methods of claims 1-14. 20

Images