Image measurement method and image processing apparatus

Abstract

The present invention provides an image measurement method and image processing apparatus that ensure the stability and accuracy of line segment measurement even for images with high variability or complex backgrounds. [Solution] The image measurement method includes acquiring an input image and multiple reference line segments within the input image; encoding the input image into a sequence vector; aligning the sequence vector and the position encoded vector with a first input vector; generating a second input vector based on the first input vector using a transformer encoder; generating a predicted offset result corresponding to each line segment to be measured based on the second input vector and query information using a transformer decoder; and generating a measurement result for each line segment to be measured based on the predicted offset result for each line segment to be measured.

Description

Technical Field

[0001] The present invention relates to an image processing mechanism, and particularly to an image measurement method and an image processing apparatus.

Background Art

[0002] In conventional computer vision technology, in order to obtain a line segment formed by the position of a measurement target in an image, feature detection and matching techniques are usually used.

Summary of the Invention

Problems to be Solved by the Invention

[0003] However, in these methods, when processing images with high variability or images with complex backgrounds, it is often difficult to ensure stability and accuracy.

[0004] Also, when it is necessary to adjust or change the measurement target, usually, redesigning and optimizing the detection algorithm takes a lot of time and resources, significantly increasing the development cost and reducing the efficiency.

Means for Solving the Problems

[0005] In view of this, the present invention provides an image measurement method and an image processing apparatus that can be used to solve the above technical problems.

[0006] Embodiments of the present invention provide an image measurement method performed by an image processing apparatus, comprising: acquiring an input image and a plurality of reference line segments within the input image, each of the reference line segments having a corresponding first endpoint position and a second endpoint position, and the plurality of reference line segments including at least a plurality of line segments to be measured; encoding the input image into a sequence vector, and aligning the sequence vector and a position-encoded vector to a first input vector, the position-encoded vector indicating the first endpoint position and the second endpoint position corresponding to each of the reference line segments; generating a second input vector based on the first input vector using a transformer encoder; generating a predicted offset result corresponding to each of the line segments to be measured based on the second input vector and query information using a transformer decoder, the query information indicating descriptive information for each of the line segments to be measured; and generating a measurement result for each of the line segments to be measured based on the predicted offset result for each of the line segments to be measured.

[0007] Embodiments of the present invention provide an image processing apparatus including a memory circuit and a processor. The memory circuit stores program code. The processor is coupled to the memory circuit and accesses the program code to acquire an input image and a plurality of reference line segments within the input image, each of which has corresponding first and second endpoint positions, and the plurality of reference line segments includes at least a plurality of line segments to be measured; encode the input image into a sequence vector and align the sequence vector and a position-encoded vector to a first input vector, the position-encoded vector indicating the first and second endpoint positions corresponding to each of the reference line segments; generate a second input vector based on the first input vector using a transformer encoder; generate a predicted offset result corresponding to each of the line segments to be measured using a transformer decoder based on the second input vector and query information, the query information indicating descriptive information for each of the line segments to be measured; and generate a measurement result for each of the line segments to be measured based on the predicted offset result for each of the line segments to be measured. [Effects of the Invention]

[0008] Based on the above, embodiments of the present invention utilize an improved DETR architecture for line segment adjustment, enabling effective capture and analysis of line segment features and their corresponding relationships within the target image and auxiliary image. This allows the image processing device to not only learn the configuration of line segments within a single image but also automatically derive the optimal mapping position for corresponding line segments within another image, thereby achieving accurate and automated line segment transformation. This innovative method not only improves processing speed and flexibility but can also be applied to fine-tuning line segments within the same image. [Brief explanation of the drawing]

[0009] [Figure 1] Figure 1 is a schematic diagram of an imaging device according to one embodiment of the present invention. [Figure 2]Figure 2 is a flowchart of an image measurement method according to one embodiment of the present invention. [Figure 3] Figure 3 is a scenario application diagram relating to one embodiment of the present invention. [Figure 4] This is a schematic diagram showing the offset positions of the first and second endpoints of each measurement target line segment, according to Figure 3. [Figure 5] This is another scenario application diagram following Figure 3. [Modes for carrying out the invention]

[0010] Referring to Figure 1, it is a schematic diagram of an image apparatus according to one embodiment of the present invention. In different embodiments, the image processing apparatus 100 can be implemented, for example, as various smart devices and / or computer devices, but is not limited thereto.

[0011] In Figure 1, the image processing device 100 includes a memory circuit 102 and a processor 104.

[0012] The memory circuit 102 is, for example, any form of fixed or portable random access memory (RAM), read-only memory (ROM), flash memory, hard disk, or other similar device, or a combination thereof, and can be used to store multiple program codes or modules.

[0013] The processor 104 is coupled to the memory circuit 102 and may be a general-purpose processor, a dedicated processor, a conventional processor, a digital signal processor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, a controller, a microcontroller, an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), any other type of integrated circuit, a state machine, an Advanced RISC Machine (ARM) based processor, and similar products.

[0014] In one embodiment of the present invention, the processor 104 can access modules and program code recorded in the memory circuit 102 to realize the image measurement method presented in the present invention, the details of which will be described below.

[0015] Referring to Figure 2, it is a flowchart of an image measurement method according to one embodiment of the present invention. The method of this embodiment can be performed by the image processing apparatus 100 of Figure 1, and below, the details of each step in Figure 2 will be described with reference to the components shown in Figure 1. In order to better understand the concept of the embodiment of the present invention, the following will be described with reference to Figure 3. Here, Figure 3 is a scenario application diagram according to one embodiment of the present invention.

[0016] In step S210, the processor 104 acquires the input image 300 and multiple reference line segments 311-314 and 321-324 within the input image 300. Here, the reference line segments 311-314 and 321-324 include at least multiple line segments to be measured.

[0017] In the scenario of FIG. 3, the input image 300 includes a joined image formed by joining the measurement target image 31 and the auxiliary image 32. Here, the measurement target image 31 and the auxiliary image 32 may be joined vertically, for example, to form the input image 300 (i.e., the joined image). In other embodiments, the measurement target image 31 and the auxiliary image 32 may be joined by other methods to form the corresponding joined image, but it is not limited thereto.

[0018] In FIG. 3, the reference line segments 311-314, 321-324 include a plurality of measurement target line segments and a plurality of auxiliary line segments. Here, the plurality of measurement target line segments include, for example, the reference line segments 311-314 located within the measurement target image 31, and the plurality of auxiliary line segments include, for example, the reference line segments 321-324 located in the auxiliary image 32.

[0019] In one embodiment, each measurement target line segment (e.g., reference line segments 311-314) may be, for example, a first line segment randomly assigned within the measurement target image 31. In the scenario of FIG. 3, the measurement target image 31 may correspond to, for example, an image of a fin field-effect transistor (FinFET). In one embodiment, a specific part (or width) of the FinFET within the measurement target image 31 may be, for example, the target to be measured.

[0020] In one embodiment, the position and / or length of each measurement target line segment (e.g., reference line segments 311-314) may be set randomly, or may be set by the designer according to certain principles, but it is not limited thereto.

[0021] In FIG. 3, each auxiliary line segment (e.g., reference line segments 321-324) indicates, for example, a second line segment with a corresponding target length within the auxiliary image 32. In the scenario of FIG. 3, the auxiliary image 32 may correspond to, for example, an image of another FinFET, and the actual length (or width) of a specific part of this FinFET may be indicated by the corresponding auxiliary line segment (e.g., reference line segments 321-324).

[0022] In other embodiments, the number of measurement target line segments in the measurement target image 31 and the number of auxiliary line segments in the auxiliary image 32 can be set according to the needs of the designer, and is not limited to the mode shown in FIG. 3.

[0023] In one embodiment, each of the reference line segments 311-314, 321-324 has corresponding first endpoint positions (e.g., left endpoint positions) and second endpoint positions (e.g., right endpoint positions).

[0024] In step S220, the processor 104 encodes the input image 300 into the sequence vector V00, and aligns the sequence vector V00 and the position encoding vector PV into the first input vector V01.

[0025] In FIG. 3, the processor 104 may, for example, take in the input image 300 into a convolutional neural network (CNN) 33, and extract the features of the input image 300 as the sequence vector V01 by the CNN 33, but is not limited thereto.

[0026] In one embodiment, the position encoding vector PV can indicate the first endpoint position and the second endpoint position corresponding to each of the reference line segments 311-314, 321-324.

[0027] Taking the reference line segment 311 as an example, the corresponding first endpoint position and second endpoint position can be represented, for example, as the pixel coordinates of the left and right endpoints of the reference line segment 311 in the measurement target image 31. Taking the reference line segment 312 as an example, the corresponding first endpoint position and second endpoint position can be represented, for example, as the pixel coordinates of the left and right endpoints of the reference line segment 312 in the measurement target image 31. Taking the reference line segment 313 as an example, the corresponding first endpoint position and second endpoint position can be represented, for example, as the pixel coordinates of the left and right endpoints of the reference line segment 313 in the measurement target image 31. Taking the reference line segment 314 as an example, the corresponding first endpoint position and second endpoint position can be represented, for example, as the pixel coordinates of the left and right endpoints of the reference line segment 314 in the measurement target image 31.

[0028] Taking reference line segment 321 as an example, its corresponding first and second endpoint positions can be represented, for example, as the pixel coordinates of the left and right endpoints of reference line segment 321 in the auxiliary image 32. Taking reference line segment 322 as an example, its corresponding first and second endpoint positions can be represented, for example, as the pixel coordinates of the left and right endpoints of reference line segment 322 in the auxiliary image 32. Taking reference line segment 323 as an example, its corresponding first and second endpoint positions can be represented, for example, as the pixel coordinates of the left and right endpoints of reference line segment 323 in the auxiliary image 32. Taking reference line segment 324 as an example, its corresponding first and second endpoint positions can be represented, for example, as the pixel coordinates of the left and right endpoints of reference line segment 324 in the auxiliary image 32.

[0029] In this case, the position-encoded vector PV in Figure 3 can, for example, represent the pixel coordinates of the left and right endpoints of the reference line segments 311 to 314 in the measurement target image 31, and the pixel coordinates of the left and right endpoints of the reference line segments 321 to 324 in the auxiliary image 32, but is not limited to this.

[0030] In some embodiments, each of the above pixel coordinates may be represented as a corresponding vector within a position-encoded vector PV after undergoing a specific transformation, but is not limited thereto.

[0031] In one embodiment, after obtaining the sequence vector V00 and the position-encoded vector PV, the processor 104 can match the sequence vector V00 and the position-encoded vector PV to the first input vector V01.

[0032] In some embodiments, the processor 104 can align the sequence vector V00 and the position-encoded vector PV to the first input vector V01 using, for example, any conventional method (e.g., concatenation).

[0033] In step S230, the processor 104 uses the transformer encoder 34 to generate a second input vector V02 based on the first input vector V01.

[0034] In one embodiment, the transformer encoder 34 is, for example, a transformer encoder in a detection transformer (DETR) model, and its detailed operating principle can be found in prior art literature related to the DETR model and will not be described in detail here.

[0035] In some embodiments, the second input vector V02 is a feature matrix globally contextualized by, for example, a transformer encoder 34, and this data includes, but is not limited to, spatial structure and semantic information.

[0036] In step S240, the processor 104 uses the transformer decoder 35 to generate predicted offset results P01 to P04 corresponding to each line segment to be measured, based on the second input vector V02 and query information Q01 to Q04. Here, the query information Q01 to Q04 represents the descriptive information for each line segment to be measured.

[0037] In one embodiment, the descriptive information for each line segment to be measured may include a semantic embedding vector corresponding to the name of each line segment to be measured. For example, assuming that the names of the reference line segments 311 to 314 (i.e., the line segments to be measured as exemplified in Figure 3) are "Line Segment to be Measured 1", "Line Segment to be Measured 2", "Line Segment to be Measured 3", and "Line Segment to be Measured 4", the processor 104 can, for example, individually input the above names into bidirectional encoding representations from Transformers (BERT), and generate semantic embedding vectors corresponding to the names of each line segment to be measured using BERT. In this case, the semantic embedding vectors corresponding to the names of each line segment to be measured may, but are not limited to, query information Q01 to Q04.

[0038] In one embodiment, the transformer decoder 35 is, for example, a DETR model transformer decoder, and its detailed operating principle can be found in prior art documents related to the DETR model, which will not be described in detail here.

[0039] In one embodiment, the plurality of line segments to be measured (e.g., reference line segments 311 to 314) include a first line segment to be measured (e.g., any of the line segments to be measured), and the predicted offset result corresponding to the first line segment to be measured includes a first offset and a second offset, where the first offset corresponds to the first endpoint position of the first line segment to be measured, and the second offset corresponds to the second endpoint position of the first line segment to be measured.

[0040] Taking reference line segment 311 as an example, the corresponding predicted offset result P01 includes, for example, a first offset corresponding to the first endpoint position of reference line segment 311 and a second offset corresponding to the second endpoint position of reference line segment 311. Taking reference line segment 312 as an example, the corresponding predicted offset result P02 includes, for example, a first offset corresponding to the first endpoint position of reference line segment 312 and a second offset corresponding to the second endpoint position of reference line segment 312. Taking reference line segment 313 as an example, the corresponding predicted offset result P03 includes, for example, a first offset corresponding to the first endpoint position of reference line segment 313 and a second offset corresponding to the second endpoint position of reference line segment 313. Taking reference line segment 314 as an example, the corresponding predicted offset result P04 includes, for example, a first offset corresponding to the first endpoint position of reference line segment 314 and a second offset corresponding to the second endpoint position of reference line segment 314, but is not limited to this.

[0041] In one embodiment, each of the first offsets may include, for example, a vertical offset component and a horizontal offset component associated with the corresponding first endpoint position. Similarly, each of the second offsets may include, but is not limited to, a vertical offset component and a horizontal offset component associated with the corresponding second endpoint position.

[0042] In one embodiment, each component corresponding to the dashed box in Figure 3 can be understood as a whole as a variation of the DETR model. In order for the illustrated variation of the DETR model to provide the above-mentioned function (for example, generating predicted offset results P01~P04 in response to the input image 300), in the relevant training process of the DETR model in Figure 3, the designer can incorporate specially designed training data into the DETR model in Figure 3, thereby enabling the DETR model in Figure 3 to perform the corresponding learning.

[0043] In some embodiments, the training data includes, for example, other images having similar properties to the input image 300. That is, each training data includes, for example, a joined image obtained by joining two FinFET images (one of which may correspond to the image to be measured and the other to a support image), and may mark a plurality of support line segments and a plurality of line segments to be measured. The marked support line segments are, for example, line segments indicating the corresponding target length in the FinFET image (e.g., width at a specific position on the FinFET), and the marked line segments to be measured may be, for example, a plurality of randomly assigned line segments, but are not limited thereto.

[0044] Furthermore, each training data set may include offset results that allow the corresponding line segment to be offset to an appropriate position.

[0045] Based on this, after incorporating the above training data into the DETR model in Figure 3 during training, the DETR model in Figure 3 can learn how to accurately offset the line segment to be measured to the appropriate position. Based on this, when an input image with one or more line segments to be measured marked is taken into the trained DETR model in Figure 3, the DETR model in Figure 3 can predict / determine how to offset the first and second positional endpoints of the line segment to accurately measure the corresponding target length, and determine the corresponding predicted offset result, but is not limited to this.

[0046] In step S250, the processor 104 generates measurement results for each line segment to be measured (e.g., reference line segments 311-314) based on the predicted offset results P01-P04 for each line segment to be measured.

[0047] Taking the first line segment to be measured as an example again, the processor 104 can, for example, offset the first endpoint position of the first line segment to be measured according to the corresponding first offset, and offset the second endpoint position of the first line segment to be measured according to the corresponding second offset. Then, the processor 104 can determine the distance between the offset first endpoint position and the offset second endpoint position of the first line segment to be measured, and determine it as the measurement result corresponding to the first line segment to be measured.

[0048] Referring to Figure 4, it is a schematic diagram in which the first and second endpoint positions of each line segment to be measured are offset according to Figure 3.

[0049] In Figure 4, the processor 104 can offset the first endpoint position (e.g., the left endpoint position) of the reference line segment 311 according to, for example, the first offset of the predicted offset result P01, thereby forming the first endpoint position (e.g., the left endpoint position) of the reference line segment 311a. Furthermore, the processor 104 can offset the second endpoint position (e.g., the right endpoint position) of the reference line segment 311 according to, for example, the second offset of the predicted offset result P01, thereby forming the second endpoint position (e.g., the right endpoint position) of the reference line segment 311a.

[0050] Subsequently, the processor 104 can determine the distance between the first and second endpoints of the reference line segment 311a and obtain the measurement result corresponding to the reference line segment 311a (or reference line segment 311).

[0051] Furthermore, the processor 104 can, for example, offset the first endpoint position (e.g., the left endpoint position) of the reference line segment 312 according to the first offset of the predicted offset result P02, thereby forming the first endpoint position (e.g., the left endpoint position) of the reference line segment 312a. The processor 104 can also, for example, offset the second endpoint position (e.g., the right endpoint position) of the reference line segment 312 according to the second offset of the predicted offset result P02, thereby forming the second endpoint position (e.g., the right endpoint position) of the reference line segment 312a.

[0052] Subsequently, the processor 104 can determine the distance between the first and second endpoint positions of the reference line segment 312a and obtain the measurement result corresponding to the reference line segment 312a (or reference line segment 312).

[0053] Regarding reference line segments 313 and 314, the processor 104 can determine the corresponding reference line segments 313a and 314a and the corresponding measurement results in accordance with the above description, and the details of this will not be explained in detail here.

[0054] As can be seen above, embodiments of the present invention utilize an improved DETR architecture for line segment adjustment, enabling effective capture and analysis of line segment features and their corresponding relationships within the target image 31 and the auxiliary image 32. This allows the image processing device 100 to not only learn the configuration of line segments within a single image (e.g., auxiliary image 32) but also to automatically derive the optimal mapping position for corresponding line segments within another image (e.g., target image 31), thereby achieving accurate and automated line segment transformation. This innovative method not only improves processing speed and flexibility but can also be applied to fine-tuning line segments within the same image.

[0055] In the scenario of Figure 3, the input image 300 considered includes a joined image formed by joining the measurement target image 31 and the auxiliary image 32. However, in other embodiments, the method presented in the embodiments of the present invention can also be applied to input images that include only the measurement target image 31.

[0056] Referring to Figure 5, it is an application diagram of another scenario following Figure 3. In the scenario of Figure 5, the input image 500 considered includes only the image to be measured 31. In this case, the reference line segments 311-314 consist only of the line segments to be measured, and each line segment to be measured is the first line segment randomly assigned to the image to be measured 31.

[0057] Based on this, the processor 104 can perform step S220 based on the input image 500 and the reference line segments 311 to 314 within it. In step S220, the processor 104 encodes the input image 500 into a sequence vector V00 and can match the sequence vector V00 and the position encoded vector PV to the first input vector V01.

[0058] In the scenario shown in Figure 5, the position-encoded vector PV can represent only the first and second endpoint positions corresponding to each measurement target line segment (i.e., reference line segments 311-314), but is not limited to this.

[0059] Subsequently, the processor 104 can continue executing steps S230 to S240. Details of each step can be found in the relevant explanations in Figures 3 and 4 and will not be explained in detail here.

[0060] In one embodiment, the various components corresponding to the dashed boxes in Figure 5 can be understood as a single unit, a variation of the DETR model. In order for the illustrated variation of the DETR model to provide the above-mentioned functions (for example, generating predicted offset results P01 to P04 according to the input image 500), specially designed training data can be incorporated into the DETR model in Figure 5 during the relevant training process of the DETR model in Figure 5, enabling the DETR model in Figure 5 to perform the corresponding learning.

[0061] In some embodiments, the training data includes, for example, other images having similar properties to the input image 500. That is, each training data may be, for example, a FinFET image, which may have multiple line segments to be measured marked therein, and these line segments to be measured may be, for example, randomly assigned line segments, but are not limited thereto.

[0062] Furthermore, each training data set may include offset results that allow the corresponding line segment to be offset to an appropriate position.

[0063] Based on this, when the above training data is fed into the DETR model in Figure 5 during training, the DETR model in Figure 5 can learn how to accurately offset the line segment to be measured to the appropriate position. Based on this, when an input image with one or more line segments to be measured marked is fed into the trained DETR model in Figure 5, the DETR model in Figure 5 can predict / determine how to offset the first and second positional endpoints of the line segment to accurately measure the corresponding target length, and determine the corresponding predicted offset result, but is not limited to this.

[0064] It should be noted that while all of the above embodiments use FinFET images as illustrative examples, this is for illustrative purposes only and is not intended to limit possible implementations of the present invention. In scenarios where measurements need to be performed on other types of images, all methods presented in the embodiments of the present invention can operate on similar principles to achieve the required measurements.

[0065] In summary, embodiments of the present invention provide a novel end-to-end automated image measurement method. Because this image measurement method processes the entire flow from image recognition to measurement line segment mapping directly within the model, it not only eliminates cumbersome post-processing steps but also avoids the high costs associated with retraining and redesigning post-processing when needs change.

[0066] Furthermore, the DETR deformation structure of the embodiment of the present invention allows the model to flexibly adapt to various measurement tasks. Unlike conventional methods that rely on clear object definitions and feature extraction, this technology uses deep learning to naturally learn the relationships between images and measurement lines, theoretically covering all future scenarios where measurements may be required. This design not only reduces the model's dependence on specific scenarios but also improves its adaptability to new and changing tasks.

[0067] Furthermore, since the embodiments of the present invention possess a high degree of automation and versatility, the technology of the present invention can be widely applied in various practical scenarios such as industrial automation, medical image analysis, and remote sensing image processing. Once sufficient training data is acquired, the model can accurately measure new images without additional adjustments, significantly reducing operational complexity and cost, and improving work efficiency.

[0068] Although the present invention has been disclosed through embodiments as described above, this does not limit the invention. Those with ordinary skill in the art can make some changes and modifications without departing from the spirit and scope of the invention. Therefore, the scope of protection of the present invention shall be determined by the scope of the appended patent application. [Industrial applicability]

[0069] The image measurement method and image processing apparatus presented in the embodiments of the present invention can be applied to fields such as industrial automation, medical image analysis, and remote sensing image processing. [Explanation of symbols]

[0070] 100: Image processing device 102:Memory circuit 104: Processor S210~S250: Process 300, 500: Input images 31: Image to be measured 32: Susuke image 33: CNN 34: Transformer Encoder 35: Transformers Decoder 311a~314a, 311~314, 321~324: Reference line segments V00: Sequence vector V01: First input vector V02: Second input vector Q01~Q04: Query Information P01~P04: Predicted offset results PV: Position Encoded Vector

Claims

1. An image measurement method performed by an image processing device, The method involves acquiring an input image and a plurality of reference line segments within the input image, wherein each of the reference line segments has a corresponding first endpoint position and a second endpoint position, and the plurality of reference line segments include at least a plurality of line segments to be measured. The input image is encoded into a sequence vector, and the sequence vector and the position encoded vector are matched to a first input vector, wherein the position encoded vector indicates the first endpoint position and the second endpoint position corresponding to each of the reference line segments. A transformer encoder is used to generate a second input vector based on the first input vector, Using a transformer decoder, predictive offset results corresponding to each of the line segments to be measured are generated based on the second input vector and query information indicating the descriptive information of each line segment to be measured. The measurement results for each of the line segments to be measured are generated based on the predicted offset results for each of the line segments to be measured, Image measurement method including

2. The input image includes the image to be measured, the plurality of reference line segments include only the plurality of line segments to be measured, and each of the line segments to be measured is a first line segment randomly assigned within the image to be measured. The method according to claim 1.

3. The input image includes a joined image obtained by joining a measurement target image and an auxiliary image, the plurality of reference line segments include the plurality of measurement target line segments and the plurality of auxiliary line segments, each of the measurement target line segments is a first line segment randomly assigned within the measurement target image, and each of the auxiliary line segments is a second line segment indicating the corresponding target length within the auxiliary image. The method according to claim 1.

4. Encoding the input image into the sequence vector is, This includes inputting the aforementioned input image into a convolutional neural network and extracting the features of the aforementioned input image as the aforementioned sequence vector using the convolutional neural network. The method according to claim 1.

5. The descriptive information for each of the line segments to be measured includes a semantic embedding vector corresponding to the name of each line segment. The method according to claim 1.

6. The plurality of line segments to be measured include a first line segment to be measured, and the predicted offset result corresponding to the first line segment to be measured includes a first offset and a second offset. The first offset corresponds to the first endpoint position of the first line segment to be measured, and the second offset corresponds to the second endpoint position of the first line segment to be measured. The method according to claim 1.

7. To generate the measurement result for each of the line segments to be measured based on the predicted offset result for each of the line segments to be measured is to The first endpoint position of the first line segment to be measured is offset according to the first offset, and the second endpoint position of the first line segment to be measured is offset according to the second offset, This includes determining the distance between the offset first endpoint position and the offset second endpoint position of the first line segment to be measured, and determining the distance as the measurement result corresponding to the first line segment to be measured. The method according to claim 6.

8. A memory circuit that stores program code, The memory circuit is connected to the program code, The method involves acquiring an input image and a plurality of reference line segments within the input image, wherein each of the reference line segments has a corresponding first endpoint position and a second endpoint position, and the plurality of reference line segments include at least a plurality of line segments to be measured. The input image is encoded into a sequence vector, and the sequence vector and the position encoded vector are matched to a first input vector, wherein the position encoded vector indicates the first endpoint position and the second endpoint position corresponding to each of the reference line segments. A transformer encoder is used to generate a second input vector based on the first input vector, Using a transformer decoder, predictive offset results corresponding to each of the line segments to be measured are generated based on the second input vector and query information indicating the descriptive information of each line segment to be measured. The measurement results for each of the line segments to be measured are generated based on the predicted offset results for each of the line segments to be measured, A processor that executes, Image processing device including

Images