Object processing methods, devices, electronic equipment, and media based on conflict verification of engineering drawings

Abstract

This disclosure presents an object processing method, apparatus, electronic device, and medium based on engineering drawing conflict verification. One specific implementation of the method includes: performing text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings to construct a drawing constraint database; obtaining candidate field values ​​and source annotation information; based on the source annotation information corresponding to the target field and the drawing constraint database, performing data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field to obtain a conflict verification result; obtaining a drawing verification score; and controlling the object processing equipment to execute the object processing task. This implementation can reduce material waste during object processing.

Description

Technical Field

[0001] Embodiments of this disclosure relate to the field of computer technology, and more specifically to object processing methods, apparatus, electronic devices, and media based on conflict verification of engineering drawings. Background Technology

[0002] In engineering design fields such as architecture and structure, it is necessary to process corresponding objects based on engineering drawings. Currently, when processing objects based on engineering drawings, the common approach is to directly convert the global description document and engineering drawings into structured data using a large language model, and then process the objects that need to be processed based on the structured data.

[0003] However, when processing objects according to engineering drawings using the above method, the following technical problems often arise:

[0004] When there is a conflict between the data in the engineering drawings and the global description document, the large language model often has difficulty in determining the correctness of the data, which can easily lead to low reliability of the data output by the large language model. This, in turn, results in low reliability of the generated structured data, causing errors in the processing of objects based on the structured data, and ultimately leading to waste of processing materials.

[0005] The information disclosed in this background section is only intended to enhance the understanding of the background of the present disclosure concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0007] Some embodiments of this disclosure provide object processing methods, apparatuses, electronic devices, and computer-readable media based on engineering drawing conflict verification to solve one or more of the technical problems mentioned in the background section above.

[0008] In a first aspect, some embodiments of this disclosure provide an object processing method based on engineering drawing conflict verification. The method includes: in response to receiving a processing signal sent by an object processing device, performing text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings to construct a drawing constraint database; based on preset target fields, performing data extraction and text association processing on pre-acquired local data of the engineering drawings to obtain candidate field values ​​and source annotation information; for each of the target fields, performing data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field based on the source annotation information corresponding to the target field and the drawing constraint database to obtain a conflict verification result; performing feasibility entropy quantification processing on the obtained conflict verification results to obtain a drawing verification score; and in response to determining that the drawing verification score is greater than a preset score threshold, controlling the object processing device to execute an object processing task based on the drawing constraint database, the target fields, the candidate field values, and the conflict verification results.

[0009] Secondly, some embodiments of this disclosure provide an object processing apparatus based on engineering drawing conflict verification. The apparatus includes: a construction unit configured to, in response to receiving a processing signal sent by an object processing device, perform text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings to construct a drawing constraint database; a first processing unit configured to, based on preset target fields, perform data extraction and text association processing on pre-acquired local data of engineering drawings to obtain candidate field values ​​and source annotation information; a second processing unit configured to, for each of the target fields, perform data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field based on the source annotation information corresponding to the target field and the drawing constraint database to obtain a conflict verification result; a quantization unit configured to perform feasibility entropy quantization processing on the obtained conflict verification results to obtain a drawing verification score; and a control unit configured to, in response to determining that the drawing verification score is greater than a preset score threshold, control the object processing device to perform an object processing task based on the drawing constraint database, the target fields, the candidate field values, and the conflict verification results.

[0010] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0011] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.

[0012] The above embodiments of this disclosure have the following beneficial effects: The object processing method based on engineering drawing conflict verification according to some embodiments of this disclosure can reduce material waste during object processing. Specifically, the reason for material waste during object processing is that when data in the engineering drawing conflicts with the global description document, the large language model often has difficulty determining the correctness of the data, easily leading to low reliability of the data output by the large language model, and consequently, low reliability of the generated structured data. This results in processing errors when processing objects based on the structured data, leading to material waste. Based on this, the object processing method based on engineering drawing conflict verification according to some embodiments of this disclosure firstly, in response to receiving a processing signal sent by the object processing device, performs text extraction and constraint recognition processing on the pre-acquired global engineering drawing file to construct a drawing constraint database. Thus, constraint information can be extracted from the global engineering drawing file to construct the drawing constraint database. Secondly, based on preset target fields, data extraction and text association processing are performed on the pre-acquired local engineering drawing data to obtain candidate field values ​​and source annotation information. Thus, candidate field values ​​of the above-mentioned target fields can be extracted from the local engineering drawing data, and the source of the candidate field values ​​can be obtained. Then, for each of the aforementioned target fields, based on the source annotation information corresponding to the target field and the aforementioned drawing constraint database, data constraint matching and conflict verification processing are performed on the candidate field values ​​corresponding to the target field to obtain conflict verification results. Thus, the candidate field values ​​of each target field can be compared with the data in the drawing constraint database. When a candidate field value conflicts with the data in the drawing constraint database, the conflict verification result can be determined based on the source annotation information corresponding to the candidate field value. Then, the obtained conflict verification results are subjected to feasibility entropy quantification processing to obtain a drawing verification score. Thus, the feasibility of the engineering drawings can be evaluated based on the conflict verification results to obtain a drawing verification score. Finally, in response to determining that the drawing verification score is greater than a preset score threshold, the object processing equipment is controlled to execute the object processing task based on the aforementioned drawing constraint database, the aforementioned target fields, the aforementioned candidate field values, and the aforementioned conflict verification results. Therefore, when the reliability of the engineering drawings is high, the conflicting candidate field values ​​can be corrected using the conflict verification results and the drawing constraint database, and then the object processing task can be executed based on the corrected data.Because the candidate field values ​​of each target field can be extracted from the local data of the engineering drawings first, and the source annotation information of the candidate field values ​​can be marked, and when there is a conflict between the candidate field values ​​and the data in the drawing constraint database, the candidate field values ​​can be corrected according to the source annotation information and the conflict verification results, and then the object can be processed using the corrected data. Therefore, the reliability of each generated candidate field value can be improved, and the waste of processing materials caused by object processing errors can be reduced. Attached Figure Description

[0013] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0014] Figure 1 This is a flowchart of some embodiments of the object processing method based on engineering drawing conflict verification according to this disclosure;

[0015] Figure 2 These are schematic diagrams of some embodiments of an object processing apparatus based on engineering drawing conflict verification according to the present disclosure;

[0016] Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0017] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0018] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0019] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0020] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0021] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0022] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0023] Figure 1 A flow 100 of some embodiments of an object processing method based on engineering drawing conflict verification according to this disclosure is shown. This object processing method based on engineering drawing conflict verification includes the following steps:

[0024] Step 101: In response to receiving a processing signal sent by the object processing equipment, perform text extraction and constraint recognition processing on the pre-acquired global file of engineering drawings to construct a drawing constraint database.

[0025] In some embodiments, the execution subject (e.g., a computing device) of the object processing method based on engineering drawing conflict verification can, in response to receiving a processing signal sent by the object processing device, perform text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings to construct a drawing constraint database.

[0026] The aforementioned object processing equipment can be any device capable of cutting or processing objects. For example, it can be a CNC machine tool. The aforementioned processing signal can be a signal indicating that processing of the object needs to begin. The aforementioned global engineering drawing file can be a CAD file representing the general requirements of the project. The aforementioned CAD file can be a file exported using CAD software (Computer-Aided Design). The aforementioned global engineering drawing file can include various drawing frames and text paragraphs. For example, the aforementioned global engineering drawing file can include, but is not limited to, "All wall heights are uniformly 3 meters". The aforementioned drawing constraint database can be a database storing various constraint fields and their values. Each constraint field can be a preset field. Each constraint field value can be the field value corresponding to the constraint field. There is a one-to-one correspondence between constraint fields and constraint field values. Each constraint field value corresponds to a unit. For example, when the global engineering drawing file is "All wall heights are uniformly 3 meters", the constraint field can be "wall height", the constraint field value can be "3", and the unit corresponding to the constraint field value can be "meters". The aforementioned execution entity can be a server.

[0027] In some optional implementations of certain embodiments, the aforementioned execution entity may perform text extraction and constraint recognition processing on the pre-acquired global engineering drawing file through the following steps to construct a drawing constraint database:

[0028] The first step is to perform frame recognition processing on the aforementioned global engineering drawing file to obtain global frame data for each frame. Each global frame data point can be a number representing the corresponding global frame. These global frames can be the frames within the aforementioned global engineering drawing file.

[0029] Each of the aforementioned global frame data sets corresponds to a frame name and boundary coordinate data. The frame name can be the name corresponding to the global frame data. The boundary coordinate data can be a matrix used to represent the position of the global frame in the aforementioned engineering drawing global file. For example, the boundary coordinate data can be [[1,10][5,3]], which can represent that the coordinates of the lower left vertex of the global frame in the aforementioned engineering drawing global file are (1,10), and the coordinates of the upper right vertex of the global frame in the aforementioned engineering drawing global file are (5,3).

[0030] In practice, the aforementioned execution entity can use a drawing frame data extraction tool to extract the drawing frame number, block name, and boundary range data of each drawing frame from the global file of the aforementioned engineering drawings. This drawing frame data extraction tool can be any tool capable of extracting geometric and attribute data of drawing frames from CAD files. For example, the drawing frame data extraction tool could be the Teigha SDK. The block name can be the name set by the technician during the drawing frame's generation. The boundary range data can be a matrix representing the position of the drawing frame within the global file of the aforementioned engineering drawings.

[0031] Then, the extracted frame numbers can be identified as individual global frame data. For each global frame data point, the corresponding block name can be identified as the frame name. Finally, the corresponding boundary range data can be identified as the corresponding boundary coordinate data.

[0032] The second step involves filtering the global frame data based on the frame names corresponding to each global frame data to obtain the target frame data.

[0033] Each of the target frame data mentioned above can be global frame data obtained after filtering.

[0034] In practice, for each of the aforementioned global frame data, in response to determining that the frame name corresponding to the global frame data contains a specific string, the aforementioned global frame data can be identified as the target frame data. The specific string can be a string pre-defined by a technician. For example, the specific string can be, but is not limited to, "General Instructions," "Unified Practice," or "General Requirements."

[0035] The third step involves extracting text content from the global engineering drawing file based on the boundary coordinates of the target drawing frames. Each image text data point can be the text extracted from the global engineering drawing file.

[0036] In practice, firstly, the aforementioned executing entity can use a data conversion tool to convert the aforementioned global engineering drawing file into an image as an engineering drawing image. This data conversion tool can be any tool capable of converting CAD files into images. For example, the data conversion tool could be Zamzar. Secondly, for each boundary coordinate data corresponding to the aforementioned target drawing frame data, firstly, the boundary coordinate data can be converted into slice range data by taking the two coordinates represented by the boundary coordinate data and proceeding in order from the smallest Y-coordinate to the largest Y-coordinate, and from the smallest X-coordinate to the largest X-coordinate. The slice range can be data used to represent the drawing frame range. For example, when the boundary coordinate data is [[1,10][5,3]], the slice range data can be [3:10,1:5].

[0037] Then, image regions corresponding to the aforementioned boundary coordinate data in the engineering drawing image can be segmented using data slicing techniques to obtain image region data. The data slicing technique can be any technique capable of segmenting a specified region in an image. For example, the data slicing technique can be NumPy array slicing.

[0038] Then, text in the aforementioned image region data can be identified using character recognition technology as image text data. This character recognition technology can be optical character recognition.

[0039] The fourth step involves performing text filtering on the aforementioned image text data based on preset target strings, resulting in target image text data. Each target image text data can be image text data containing any one of the aforementioned target strings. Each target string can be a string pre-defined by a technician. The aforementioned target strings may include, but are not limited to, "should," "must," and "all."

[0040] In practice, for each of the above image text data, in response to determining that the above image text data contains any one of the above target strings, the above image text data can be identified as target image text data.

[0041] The fifth step involves performing constraint recognition processing on the aforementioned target image text data to obtain constraint structured data. This constraint structured data can be the structured data extracted from the aforementioned target image text data. The constraint structured data can include various constraint fields and their values. Each constraint field corresponds to one constraint field value. Each constraint field and its value has a corresponding unit.

[0042] Step 6: Based on the above-mentioned structured constraint data, construct a drawing constraint database.

[0043] In practice, for each constraint field in the aforementioned structured constraint data, the constraint field and its corresponding value can be stored as key-value pairs in a preset database for updating the preset database. Finally, the updated preset database can be designated as the drawing constraint database.

[0044] In some optional implementations of certain embodiments, the aforementioned execution entity may perform constraint recognition processing on the aforementioned target image text data through the following steps to obtain constraint structured data:

[0045] For each of the target image text data points mentioned above, perform the following steps:

[0046] The first step is to determine each image field corresponding to the target image text data based on the preset field data table and the frame name corresponding to the target image text data.

[0047] The aforementioned field data table can be a table recording the various fields that need to be identified. The aforementioned field data table can include various fields. Each field in the aforementioned field data table corresponds to a unit. Each of the aforementioned image fields can be a field in the aforementioned field data table that contains the frame name corresponding to the aforementioned target image text data.

[0048] In practice, for each field in the aforementioned field data table, the fields containing the frame name corresponding to the target image text data can be designated as image fields. For example, when the frame name corresponding to the target image text data is "East Wall," the aforementioned image fields can be the fields in the aforementioned field data table that contain "East Wall."

[0049] The second step involves performing field value matching on the target image text data based on the aforementioned image fields to obtain key-value pairs of each image.

[0050] Each image key-value pair in the aforementioned image key-value pair data can be a key-value pair extracted from the aforementioned target image text data. Each image key-value pair data includes an image field and an image field value. The image field value can be the field value corresponding to the image field.

[0051] In practice, for each of the aforementioned image fields, firstly, the preset field value determination information, the image field itself, the corresponding unit, and the target image text data can be input into a large language model to obtain the output data of the large language model as the image field value. The field value determination information can be used to prompt the large language model to extract the corresponding field value from the target image text data. For example, when the image field is "East Wall Height" and the unit is "meters," the field value determination information could be "Please directly output the value of the East Wall height in meters based on the target image text data." The large language model can be DeepSeek. Secondly, the image field and its value can be combined into key-value pairs as image key-value pair data.

[0052] The third step involves performing structured cleaning on the aforementioned image key-value pairs to obtain cleaned key-value pairs. Each of these cleaned key-value pairs can be image key-value pairs that have undergone data cleaning.

[0053] In practice, for each of the aforementioned image key-value pairs, and for the image field values ​​included in the image key-value pair data, special characters in the image field values ​​can be removed using regular expressions to clean the image key-value pair data, resulting in cleaned image key-value pair data. The special characters mentioned above can be Unicode characters other than numbers, English letters, and Chinese characters.

[0054] The fourth step involves normalizing the cleaned key-value pairs to obtain structured image-text data. This structured image-text data can be composed of individual image-text key-value pairs. Each image-text key-value pair can be the processed cleaned key-value pair data. Each image-text key-value pair can include individual image-text key-value pairs. Each image-text key-value pair includes an image-text field and an image-text field value. The image-text field can be any field in the image-text key-value pair data. The image-text field value can be the field value corresponding to the image-text field. Both the image-text field and the image-text field value have corresponding units.

[0055] In practice, for each of the aforementioned cleaned key-value pairs, in response to determining that a value exists in the image field value included in the cleaned key-value pair, the following steps can be performed on the image field and image field value included in the cleaned key-value pair:

[0056] First, the image field included in the aforementioned cleansing key-value pair data can be identified as an image text field. Second, for each preset string group, in response to determining that the image field contains any preset string from that preset string group, the unit corresponding to the image field can be identified as the unit to be converted. Each preset string in each preset string group can be a pre-defined string. For example, a preset string could be "table". Each preset string group corresponds to a preset unit. The preset unit can represent the unit that the image field containing any preset string from the preset string group should correspond to. For example, the preset unit could be "cm".

[0057] Then, in response to determining that the unit to be converted is different from the preset unit corresponding to the preset string group, the image field value included in the cleaned key-value pair data, the unit to be converted, and the preset unit can be input into a unit conversion function to obtain the data output by the unit conversion function as the image text field value. The unit conversion function can be a function capable of performing unit conversion on data. For example, the unit conversion function can be the CONVERT function. In response to determining that the unit to be converted is the same as the unit to be converted, the image field value included in the cleaned key-value pair data can be determined as the image text field value. Then, the image text field and the obtained image text field value can be combined into key-value pair data as image text key-value pair data.

[0058] Then, in response to determining that no numerical value exists in the image field values ​​included in the aforementioned cleaned key-value pair data, the image field and image field values ​​included in the aforementioned cleaned key-value pair data can be respectively determined as image text field and image text field value. Then, the aforementioned image text field and the aforementioned image text field value can be combined into key-value pair data as image text key-value pair data.

[0059] Finally, the obtained image-text key-value pairs can be combined into image-text structured data.

[0060] The fifth step involves performing logical conflict verification and conflict resolution on the above image and text structured data to obtain constrained structured data.

[0061] In practice, for each image text key-value pair in the image text structured data, firstly, in response to determining that the image text key-value pair satisfies a preset key-value pair condition, the image text field included in the image text key-value pair data can be determined as the image text field to be processed. The preset key-value pair condition can be that the value of the image text field included in the image text key-value pair data is numerical, and the image text field included in the image text key-value pair data contains any one of the preset maximum / minimum value strings. Each of the preset maximum / minimum value strings can be a pre-defined string. For example, the preset maximum / minimum value strings can include, but are not limited to: "maximum", "highest", and "longest".

[0062] Then, the preset semantically opposite information, the aforementioned image text field to be processed, and the various image text fields included in the aforementioned image text structured data can be sent to the target technical terminal. The target technical terminal can be a terminal used by a technician. The semantically opposite information can be used to prompt the technician and output image text fields with semantically opposite meanings to the aforementioned image text field to be processed. For example, the semantically opposite information could be "Please return the image text field with semantically opposite meanings to the aforementioned image text field to be processed." Then, the image text field sent by the target technical terminal can be received as the opposite image text field. For example, when the image text field to be processed is "Highest value of the east wall," the opposite image text field with semantically opposite meanings to "Highest value of the east wall" could be "Lowest value of the east wall."

[0063] Then, the image text field value corresponding to the aforementioned image text field to be processed can be determined as the maximum field value. The image text field value corresponding to the aforementioned opposite image text field can be determined as the minimum field value. In response to determining that the aforementioned maximum field value is less than the aforementioned minimum field value, the aforementioned minimum field value can be determined as the image text field value corresponding to the aforementioned image text field to be processed, and the aforementioned maximum field value can be determined as the image text field value corresponding to the aforementioned opposite image text field, so as to update the image text structured data.

[0064] For example, when the image text field value of the image text field "Highest value of the East Wall" is 10, and the image text field value of the image text field "Lowest value of the East Wall" is 15, 15 can be determined as the field value of "Highest value of the East Wall," and 10 can be determined as the field value of "Lowest value of the East Wall." Then, for each image text key-value pair in the updated image text structured data, the image text field and image text field value in the above image text key-value pair data can be determined as the constraint field and constraint field value, respectively. Finally, the determined constraint fields and the determined constraint field values ​​can be combined to form constrained structured data.

[0065] Step 102: Based on the preset target fields, perform data extraction and text association processing on the pre-acquired partial data of the engineering drawings to obtain the values ​​of each candidate field and the annotation information of each source.

[0066] In some embodiments, the aforementioned execution entity may perform data extraction and text association processing on the pre-acquired partial data of engineering drawings based on preset target fields to obtain the values ​​of each candidate field and the annotation information of each source.

[0067] Each of the target fields mentioned above can be a field that needs to be extracted from the partial data of the aforementioned engineering drawings. The partial data of the engineering drawings can be the CAD file corresponding to a specific part of the engineering drawing. For example, the partial data of the engineering drawings can be the CAD file corresponding to the floor plan of the first floor of a building. Each of the candidate field values ​​mentioned above can be the field value corresponding to the target field. Each of the source annotation information mentioned above can be a label that can characterize the source of the candidate field value. For example, the source annotation information can be "explicit," "missing," or "predicted."

[0068] In addressing the technical problems mentioned above, and considering the application scenario in a mechanical hydraulic valve block processing workshop, the following technical issues often arise: When using a large language model to extract field values ​​from engineering drawings, the large language model needs to traverse the engineering drawing for each field, resulting in significant computational resource consumption during model traversal. Furthermore, the large language model is susceptible to interference from other parts of the engineering drawing, leading to lower accuracy of the extracted field values. This, in turn, can cause processing errors when processing objects based on structured data, resulting in material waste. Solving these factors can reduce material waste. Given the following requirements for this application scenario: the workshop's industrial control equipment has limited performance but needs to respond quickly to real-time processing tasks for various drawings, making it extremely sensitive to computational resource consumption; simultaneously, hydraulic valve blocks typically use high-strength alloy steel, resulting in high material costs. Material waste due to processing errors can easily lead to substantial cost losses. Therefore, to minimize material waste, we have decided to adopt the following solution:

[0069] In some optional implementations of certain embodiments, the aforementioned execution entity may perform data extraction and text association processing on pre-acquired partial data of engineering drawings based on preset target fields through the following steps to obtain the values ​​of each candidate field and the annotation information of each source:

[0070] The first step involves performing region identification processing on the aforementioned partial data of the engineering drawing to obtain bounding box data for each region. Each bounding box data point represents the location of a rectangular region identified from the partial drawing image data. For example, the bounding box data could be [1,3,5,10], representing the coordinates of the top-left corner of the rectangular region as (1,3) and the bottom-right corner as (5,10). The partial drawing image data can be an image obtained after data conversion of the aforementioned partial data of the engineering drawing. Each bounding box data point has a corresponding number and category label. The number is a unique identifier for the bounding box data. The category label represents the category of the image region corresponding to the bounding box data. For example, the category label could be "text," "graphic," or "leader line."

[0071] In practice, firstly, the aforementioned data conversion tools can be used to convert the partial data of the engineering drawings into images, serving as partial drawing image data. Secondly, the partial drawing image data can be input into a pre-trained region recognition model to obtain bounding box data for each region. This region recognition model can be a neural network model that takes the partial drawing image data as input and outputs the bounding box data for each region and the corresponding category label for each bounding box. For example, the region recognition model can be a pre-trained YOLOv8 model. This pre-training process involves fine-tuning the YOLOv8 model using the labeled partial drawing image dataset and the cross-entropy loss function. Finally, the obtained bounding box data can be numbered. For example, the first bounding box data is numbered "1", the second bounding box data is numbered "2", and so on.

[0072] The second step involves performing text recognition and feature extraction on the image regions corresponding to the aforementioned partial data of the engineering drawings and the bounding box data of each region, thereby obtaining the image region data.

[0073] Each of the aforementioned image region data can be the data corresponding to an image region in the local image data. Each of the aforementioned image region data can include region bounding box data, category labels, region text data, and region visual vectors. The aforementioned local image data can be the image corresponding to the local data of the aforementioned engineering drawing. The aforementioned region text data can be the text content within the image region represented by the image region data. The aforementioned region visual vectors can be the feature vectors corresponding to the image region represented by the image region data.

[0074] In practice, firstly, the aforementioned data conversion tool can be used to convert the partial data of the engineering drawings into images as partial image data.

[0075] Then, for each region bounding box in the aforementioned region bounding box data, the region bounding box data can be input into an image segmentation function to crop the image region corresponding to the region bounding box data from the aforementioned local image data as the target image region. The image segmentation function can be any function capable of cropping a specified region from an image. For example, the image segmentation function could be the `crop()` function.

[0076] Then, in response to determining that the category label corresponding to the aforementioned region bounding box data is "text", the text in the aforementioned target image region can be identified as region text data using the aforementioned text recognition technology. Then, feature extraction technology can be used to extract features from the aforementioned target image region to obtain the feature vector corresponding to the aforementioned target image region as the region visual vector. The aforementioned feature extraction technology can be any technology capable of extracting feature vectors from an image. For example, the aforementioned feature extraction technology can be a histogram of oriented gradients (HAR). Finally, the aforementioned region bounding box data, the category label corresponding to the aforementioned region bounding box data, the aforementioned region text data, and the aforementioned region visual vector can be combined to form image region data.

[0077] Then, in response to determining that the category label corresponding to the aforementioned region bounding box data is not "text", null values ​​can be identified as region text data corresponding to the aforementioned region bounding box data. The aforementioned feature extraction techniques can be used to perform feature extraction processing on the aforementioned target image region, obtaining the feature vector corresponding to the aforementioned target image region as the region visual vector. Finally, the aforementioned region bounding box data, the category label corresponding to the aforementioned region bounding box data, the aforementioned region text data, and the aforementioned region visual vector can be combined into image region data.

[0078] Third, for each of the target fields mentioned above, perform the following steps:

[0079] The first sub-step involves performing text feature embedding processing on the target field to obtain a text query vector. This text query vector can be a feature vector corresponding to the target field. In practice, the executing entity can use text embedding techniques to perform text feature embedding processing on the target field to obtain the text query vector. These text embedding techniques can be those capable of converting text into feature vectors. For example, the text embedding technique could be TF-IDF (term frequency-inverse document frequency).

[0080] The second sub-step involves performing the following steps for each of the aforementioned image region data:

[0081] Sub-step one involves performing text feature embedding processing on the regional text data included in the aforementioned image region data to obtain a regional text vector. This regional text vector can be a feature vector corresponding to the aforementioned regional text data. In practice, the executing entity can use the aforementioned text embedding technique to perform text feature embedding processing on the regional text data included in the aforementioned image region data to obtain a regional text vector.

[0082] Sub-step two involves comparing the similarity between the text query vector and the region text vector to obtain the text similarity score. This text similarity score can be the cosine similarity between the text query vector and the region text vector.

[0083] Sub-step three involves performing feature alignment processing on the aforementioned text query vector and the region visual vectors included in the aforementioned image region data to obtain aligned text vectors and aligned visual vectors. The aligned text vector can be a text query vector mapped to a preset semantic space. The aligned visual vector can be a region visual vector mapped to the aforementioned semantic space. The semantic space can be the vector space corresponding to a pre-trained vector alignment model.

[0084] In practice, the aforementioned execution entity can input the aforementioned text query vector and region visual vector into a pre-trained vector alignment model to obtain aligned text vectors and aligned visual vectors. The vector alignment model can be a neural network model that takes the text query vector and region visual vector as input and outputs the aligned text vector and aligned visual vector. This vector alignment model can include two multilayer perceptrons. The first multilayer perceptron can take the text query vector as input and output the aligned text vector. The second multilayer perceptron can take the region visual vector as input and output the aligned visual vector.

[0085] Sub-step four involves comparing the similarity between the aligned text vector and the aligned visual vector to obtain a visual similarity score. This visual similarity score can be the cosine similarity between the aligned text vector and the aligned visual vector.

[0086] Sub-step five involves performing feature weighting on the aforementioned text similarity and visual similarity to obtain a comprehensive similarity. The comprehensive similarity can be a numerical value obtained by weighting the aforementioned text similarity and visual similarity.

[0087] In practice, the product of the first preset weight and the aforementioned text similarity can be used to determine the first similarity weight data. The first preset weight can be a pre-set value, such as 0.7. The product of the second preset weight and the aforementioned visual similarity can be used to determine the second similarity weight data. The second preset weight can also be a pre-set value, such as 0.3. Finally, the sum of the first and second similarity weight data can be used to determine the comprehensive similarity.

[0088] The third sub-step involves performing target region segmentation on the obtained comprehensive similarity scores and the aforementioned image region data to obtain local image data. This local image data can be an image region cropped from the image data corresponding to the local data of the engineering drawing.

[0089] In practice, the comprehensive similarity score with the highest value among the aforementioned comprehensive similarities can be determined as the target comprehensive similarity score. Secondly, the image region data corresponding to the target comprehensive similarity score can be determined as the target image region data. Then, the bounding box data of the regions included in the target image region data can be determined as the target region bounding box data. Next, the aforementioned data conversion tool can be used to convert the aforementioned partial data of the engineering drawing into an image as the engineering drawing image data. Then, the aforementioned target region bounding box data can be input into the aforementioned image segmentation function to crop the image region corresponding to the target region bounding box data from the aforementioned engineering drawing image data as partial image data.

[0090] The fourth sub-step involves extracting candidate data from the aforementioned local image data based on preset field value extraction information. This yields candidate field values ​​corresponding to the target field. The extracted field value information can be used to prompt the large language model to output the target field value. The candidate field values ​​can be field values ​​output by the large language model that correspond to the target field. Each candidate field value has a corresponding confidence level. For example, the extracted field value information could be: "Based on the local image data, if there is information explicitly labeled with the target field in the local image data, please output that information; if there is no label in the local image data, please output null; the output value should follow the format {value, confidence level (0-1)}".

[0091] In practice, the aforementioned execution entity can input the extracted information of the aforementioned field values ​​and the aforementioned local image data into the aforementioned large language model, obtain the data output by the aforementioned large language model as candidate field values, and obtain the decimal of the output of the aforementioned candidate field values ​​as the confidence level of the aforementioned candidate field values.

[0092] The fifth sub-step involves determining the source annotation information corresponding to the target field based on the aforementioned candidate field values. This source annotation information can be a label that characterizes the source of the candidate field value. For example, the source annotation information could be "explicit," "missing," or "predicted."

[0093] In practice, firstly, in response to determining that the aforementioned candidate field values ​​meet the first preset data condition, the preset explicit annotation information is determined as the source annotation information corresponding to the aforementioned target field. The first preset data condition can be that the aforementioned candidate field value is not "null" and the confidence level corresponding to the aforementioned candidate field value is greater than a preset confidence threshold. The preset confidence threshold can be a pre-set value. Here, the specific setting of the preset confidence threshold is not limited. The explicit annotation information can be "explicit".

[0094] Secondly, in response to determining that the above candidate field values ​​satisfy the second preset data condition, the preset missing annotation information is determined as the source annotation information corresponding to the above target field. Here, the above second preset data condition can be that the above candidate field value is "null". The above missing annotation information can be "missing".

[0095] Then, in response to determining that the above candidate field values ​​satisfy the third preset data condition, the preset prediction annotation information is determined as the source annotation information corresponding to the above target field. The third preset data condition can be that the above candidate field value is not "null" and the confidence level corresponding to the above candidate field value is less than or equal to the above preset confidence threshold. The above prediction annotation information can be "prediction".

[0096] The above technical solution and its related content, combined with step 105, serve as an inventive point of this disclosure, solving the problem of "waste of processing materials." Factors leading to waste of processing materials are often as follows: When using a large language model to extract field values ​​from engineering drawings, the large language model needs to traverse the engineering drawing for each field, resulting in significant computational resource consumption during model traversal. Furthermore, the large language model is easily affected by interference from other parts of the engineering drawing, leading to lower accuracy of the extracted field values, which in turn causes processing errors when processing objects based on structured data, resulting in waste of processing materials. Solving these factors can reduce waste of processing materials. To achieve this effect, this disclosure first performs region identification processing on the aforementioned local data of the engineering drawing to obtain boundary box data for each region. Each boundary box data in the aforementioned boundary box data corresponds to a number and a category label. Thus, the boundary box data for each region can be obtained. Secondly, based on the local data of the aforementioned engineering drawings and the category labels corresponding to the bounding box data of each region, text recognition and feature extraction are performed on the image regions corresponding to the bounding box data of each region to obtain image region data. Each image region data includes region text data and a region visual vector. Thus, the image region data can be obtained. Then, for each target field, the following steps are performed: First, text feature embedding is performed on the target field to obtain a text query vector. Thus, the feature vector of the target field can be obtained. Next, for each image region data, the following steps are performed: Then, text feature embedding is performed on the region text data included in the image region data to obtain a region text vector. Thus, the feature vector of the region text data in the image region data can be obtained. Then, a similarity comparison is performed between the text query vector and the region text vector to obtain text similarity. Thus, the similarity between the text query vector and the region text vector can be obtained. Then, feature alignment is performed on the text query vector and the region visual vector included in the image region data to obtain aligned text vectors and aligned visual vectors. Therefore, aligned text vectors and aligned visual vectors can be obtained. Then, the similarity of the aligned text vectors and aligned visual vectors is compared to obtain visual similarity. This yields the similarity between the aligned text vectors and aligned visual vectors. Next, feature-weighted processing is applied to the text similarity and visual similarity to obtain a comprehensive similarity. This provides the similarity between the target field and each image region data. Then, based on the obtained comprehensive similarities and the image region data, the local data of the engineering drawing is segmented into target regions to obtain local image data. This yields the local image data with the highest similarity to the target field.Then, based on the preset field value extraction information, candidate data extraction processing is performed on the aforementioned local image data to obtain candidate field values ​​corresponding to the aforementioned target field, where each candidate field value corresponds to a confidence level. Thus, the field value of the aforementioned target field can be extracted from the aforementioned local image data. Finally, based on the aforementioned candidate field values, the source annotation information corresponding to the aforementioned target field is determined. Thus, source annotation information for the candidate field values ​​can be generated. Also, because when generating candidate field values ​​for the target field, only the region with the highest similarity to the target field can be identified based on the similarity between each region in the engineering drawing and the target field, the computational resources consumed during field value extraction can be reduced. Furthermore, because for a single target field, the large language model only needs to identify the region related to that target field, rather than identifying the entire local data of the engineering drawing, the interference encountered by the large language model during field value extraction can be reduced, improving the accuracy of the extracted field values. This, in turn, reduces the probability of processing errors when processing objects based on field values, and reduces material waste.

[0097] Step 103: For each target field in each target field, based on the source annotation information and drawing constraint database corresponding to the target field, perform data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field to obtain the conflict verification result.

[0098] In some embodiments, the execution entity may, for each of the target fields, perform data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field based on the source annotation information corresponding to the target field and the drawing constraint database, to obtain a conflict verification result. The conflict verification result can be information used to characterize whether there is a conflict between the candidate field value and the corresponding constraint field value in the drawing constraint database. For example, the conflict verification result can be "no conflict," "default value filling," or "conflict exists."

[0099] In some optional implementations of certain embodiments, the execution entity may perform data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field based on the source annotation information corresponding to the target field and the drawing constraint database, thereby obtaining the conflict verification result:

[0100] The first step is to query the constraint field value corresponding to the target field from the aforementioned drawing constraint database as the target constraint field value. This target constraint field value can be the constraint field value corresponding to the same constraint field stored in the drawing constraint database as the target field. In practice, the constraint field stored in the drawing constraint database that is identical to the target field can be identified as the target constraint field. Then, the constraint field value corresponding to the target constraint field can be identified as the target constraint field value.

[0101] The second step is to respond to the determination that the source annotation information corresponding to the above target field meets the preset missing annotation conditions, and to fill the candidate field values ​​corresponding to the above target field with data based on the above target constraint field values, so as to update the candidate field values ​​and determine the preset field conflict information as the conflict verification result.

[0102] The missing annotation condition mentioned above can be that the source annotation information is "missing". The field conflict information mentioned above can indicate that the candidate field value is obtained from the target constraint field value. For example, the field conflict information mentioned above can be "default value filling".

[0103] In practice, in response to determining that the value of the above target constraint field is a number or a string, the value of the above target constraint field can be determined as the candidate field value corresponding to the above target field, so as to update the candidate field value.

[0104] In response to determining that the target constraint field value is an interval, the average of the value corresponding to the left interval and the value corresponding to the right interval of the target constraint field value can be determined as the candidate field value corresponding to the target field, so as to update the candidate field value.

[0105] Third, in response to determining that the source annotation information corresponding to the above target field meets the preset explicit annotation conditions, the following steps are performed:

[0106] The first sub-step involves, in response to determining that the candidate field values ​​meet preset text data conditions, performing conflict detection processing on the candidate field values ​​and the target constraint field values ​​to obtain conflict verification results. Specifically, the explicit annotation condition can be that the source annotation information is "explicit". The text data condition can be that all candidate field values ​​are text or that the candidate field values ​​are candidate strings. The candidate strings can be strings pre-defined by technical personnel. For example, the candidate strings can be, but are not limited to, "MU15", "MU20", or "MU25".

[0107] In practice, in response to determining that all candidate field values ​​are text and that the candidate field values ​​are not candidate strings, the executing entity can use the text embedding technique to convert the candidate field values ​​and the target constraint field values ​​into feature vectors respectively. The feature vectors corresponding to the candidate field values ​​are used as candidate feature vectors, and the feature vectors corresponding to the target constraint field values ​​are used as constraint feature vectors. Then, the cosine similarity between the candidate feature vectors and the constraint feature vectors can be determined as the text field value similarity. In response to determining that the text field value similarity is greater than a preset text similarity threshold, "no conflict" can be determined as the conflict verification result. The text similarity threshold can be a pre-set value. Here, the specific setting of the text similarity threshold is not limited. Then, in response to determining that the text field value similarity is less than or equal to the text similarity threshold, "conflict exists" can be determined as the conflict verification result.

[0108] In response to determining that the candidate field value is the candidate string, the candidate field value and the target constraint field value can be input into a string comparison function, and the return value of the string comparison function can be used as the comparison result. The string comparison function can be any function capable of comparing whether two strings are identical. For example, the string comparison function could be the strcmp() function. Then, in response to determining that the comparison result indicates that the candidate field value and the target constraint field value are completely identical, "no conflict" can be determined as the conflict verification result. As an example, when the comparison result is 0, it indicates that the candidate field value and the target constraint field value are completely identical. Then, in response to determining that the comparison result cannot indicate that the candidate field value and the target constraint field value are completely identical, "conflict exists" can be determined as the conflict verification result. As an example, when the comparison result is not 0, it cannot indicate that the candidate field value and the target constraint field value are completely identical.

[0109] The second sub-step, in response to determining that the above candidate field values ​​meet the preset numerical data conditions, executes the following steps:

[0110] Sub-step one involves performing logical standardization on the aforementioned target constraint field values ​​to obtain logical constraint objects. Specifically, the aforementioned numerical data conditions require that the aforementioned candidate field values ​​are not the aforementioned candidate strings and that the aforementioned candidate field values ​​contain numerical values. The aforementioned logical constraint objects can be the aforementioned target constraint field values ​​or the intervals corresponding to the aforementioned target constraint field values.

[0111] In practice, in response to determining that the target constraint field values ​​contain only numerical values, the target constraint field values ​​can be defined as logical constraint objects, and the units corresponding to the target constraint field values ​​can be defined as the units corresponding to the logical constraint objects.

[0112] Then, in response to determining that the target constraint field value contains more than just numerical values, preset data conversion information and the target constraint field value can be sent to the target terminal. The data conversion information can be information used to prompt technicians to convert the target constraint field value into a range. The target terminal can be a terminal used by technicians. Then, the data sent by the target terminal can be received as a logical constraint object, and the unit corresponding to the target constraint field value can be determined as the unit corresponding to the logical constraint object. As an example, when the target constraint field value is greater than 10 and less than 12, the logical constraint object can be (10, 12).

[0113] Sub-step two involves performing unit conversion and conflict detection on the candidate field values ​​corresponding to the target field based on the aforementioned logical constraint object, thereby obtaining a conflict detection result. This conflict detection result can be information used to characterize whether the aforementioned logical constraint object matches the candidate field value corresponding to the target field. For example, the conflict detection result can be "match" or "not match".

[0114] In practice, firstly, in response to the fact that the unit corresponding to the target field is different from the unit corresponding to the logical constraint object, the unit corresponding to the logical constraint object can be determined as the target unit. The unit corresponding to the target field can be determined as the current unit. Then, a unit conversion function can be used to convert the candidate field value corresponding to the target field from the current unit to the target unit, obtaining the converted candidate field value as the converted field value. The unit conversion function can be any function capable of converting data units. For example, the unit conversion function could be the CONVERT function.

[0115] Then, in response to determining that the logical constraint object contains only numerical values, and in response to determining that the transformation field value is the same as the logical constraint object, "match" can be determined as a conflict detection result. In response to determining that the transformation field value is not the same as the logical constraint object, "mismatch" can be determined as a conflict detection result.

[0116] Then, in response to determining that the above logical constraint object is a range, and in response to determining that the above transformation field value belongs to the above logical constraint object, "match" can be determined as a conflict detection result. In response to determining that the above transformation field value does not belong to the above logical constraint object, "non-match" can be determined as a conflict detection result.

[0117] Sub-step three: Based on the conflict detection results, generate a conflict verification result. In practice, in response to determining that the conflict detection results meet the preset data matching conditions, "no conflict" can be determined as the conflict verification result. The data matching condition can be "matched". Then, in response to determining that the conflict detection results do not meet the data matching conditions, "conflict exists" can be determined as the conflict verification result.

[0118] Optionally, after generating the conflict verification result based on the conflict detection result, the execution entity may also perform the following steps:

[0119] In response to the determination that the above conflict verification result satisfies the preset conflict verification conditions, the following steps are performed:

[0120] The first step involves performing multi-agent conflict resolution on the target constraint field value and the corresponding candidate field value based on preset first and second prompt information, yielding first and second conflict resolution information. The conflict verification condition can be that the conflict verification result is "a conflict exists." The first prompt information can be used to inform the agent why the target constraint field value is correct. For example, the first prompt information could be "There is a conflict between the target constraint field value and the corresponding candidate field value; please output the reason why the target constraint field value is correct."

[0121] The aforementioned second prompt information can be information used to indicate to the agent that the output candidate field value corresponding to the target field is correct. The aforementioned first conflict resolution information can be text data output by the agent based on the aforementioned first prompt information. The aforementioned second conflict resolution information can be text data output by the agent based on the aforementioned second prompt information.

[0122] In practice, firstly, the executing entity can input the first prompt information, the target constraint field value, and the candidate field value corresponding to the target field into the intelligent agent, obtaining the text data output by the intelligent agent as the first conflict resolution information. The intelligent agent can be a tool capable of outputting corresponding content based on the input data. For example, the intelligent agent could be DeepSeek. Secondly, the second prompt information, the target constraint field value, and the candidate field value corresponding to the target field can be input into the intelligent agent, obtaining the text data output by the intelligent agent as the second conflict resolution information.

[0123] The second step involves intelligently parsing the first and second conflict resolution information based on preset conflict warning information, the target constraint field value, and the candidate field value corresponding to the target field, to obtain the third conflict resolution information. The conflict warning information can be used to prompt the large language model to determine which is more correct—the target constraint field value or the candidate field value corresponding to the target field—based on the first and second conflict resolution information. For example, the conflict warning information could be, "Please determine what the field value corresponding to the target field should be based on the first and second conflict resolution information." The large language model can be GPT-4. The third conflict resolution information can be the information output by the large language model.

[0124] In practice, the aforementioned executing entity can input the aforementioned conflict prompt information, the aforementioned target constraint field value, the aforementioned candidate field value corresponding to the aforementioned target field, the aforementioned first conflict resolution information, and the aforementioned second conflict resolution information into the aforementioned large language model, and obtain the information output by the aforementioned large language model as the third conflict resolution information.

[0125] The third step involves generating, based on the aforementioned third conflict resolution information, the aforementioned target constraint field value, and the aforementioned candidate field value corresponding to the target field, the field value to be adjusted corresponding to the aforementioned target field. The aforementioned field value to be adjusted can be the field value corresponding to the aforementioned target field determined according to the aforementioned third conflict resolution information.

[0126] In practice, the above-mentioned text embedding technology can be used to convert the above-mentioned third conflict resolution information, the above-mentioned target constraint field value and the above-mentioned target field corresponding candidate field value into feature vectors. The feature vector corresponding to the above-mentioned third conflict resolution information is used as the conflict resolution vector, the feature vector corresponding to the above-mentioned target constraint field value is used as the target constraint vector, and the feature vector corresponding to the above-mentioned target field corresponding candidate field value is used as the candidate vector.

[0127] Secondly, the cosine similarity between the conflict resolution vector and the target constraint vector can be determined as the first similarity. The cosine similarity between the conflict resolution vector and the candidate vector can be determined as the second similarity.

[0128] Then, in response to determining that the first similarity is greater than or equal to the second similarity, the target constraint field value can be determined as the field value to be adjusted corresponding to the target field. In response to determining that the first similarity is less than the second similarity, the candidate field value corresponding to the target field can be determined as the field value to be adjusted corresponding to the target field.

[0129] The fourth step involves adaptively adjusting the value of the field to be adjusted based on the target field, the value of the field to be adjusted, and the preset adjustment information to obtain the target field value.

[0130] The aforementioned adjustment information can serve as a prompt to technicians to adjust the value of the field to be adjusted. In practice, the executing entity can input the target field, the value of the field to be adjusted, and the adjustment information into the target terminal. Then, it can receive data sent by the target terminal as the target field value.

[0131] The fifth step is to combine the target field and its value into target key-value pairs.

[0132] Step 104: Perform feasibility entropy quantification on the obtained conflict verification results to obtain the drawing verification score.

[0133] In some embodiments, the execution entity may perform feasibility entropy quantification on the obtained conflict verification results to obtain a drawing verification score. The drawing verification score can be a numerical value used to characterize the reliability of local data in the engineering drawings.

[0134] In some optional implementations of certain embodiments, the aforementioned execution entity may perform feasibility entropy quantification on the obtained conflict verification results through the following steps to obtain the drawing verification score:

[0135] The first step, based on the conflict verification results mentioned above, is to determine the number of explicit fields, the number of default fields, and the number of conflicting fields. Specifically, the number of explicit fields can be the number of conflict verification results with a value of "no conflict". The number of default fields can be the number of conflict verification results with a value of "filled with default value". The number of conflicting fields can be the number of conflict verification results with a value of "conflict exists".

[0136] In practice, the number of conflict verification results with a value of "no conflict" can be determined as the number of explicit fields.

[0137] Then, the number of conflict verification results with the value "filled with default value" can be determined as the number of default fields.

[0138] Then, the number of conflict verification results with a value of "conflict exists" can be determined as the number of conflict fields.

[0139] The second step involves weighting the number of explicit fields, the number of default fields, and the number of conflicting fields based on preset explicit weight data, preset default weight data, and preset conflict weight data to obtain a total field weight value. This total field weight value can be the numerical value obtained by weighting the number of explicit fields, default fields, and conflicting fields. The explicit weight data can be pre-set values. For example, the explicit weight data can be 0.3. The default weight data can be pre-set values. For example, the default weight data can be 0.3. The conflict weight data can be pre-set values. For example, the conflict weight data can be 0.4.

[0140] In practice, firstly, the product of the explicitly weighted data and the explicitly number of fields can be used to determine the explicitly weighted value. Secondly, the product of the default weighted data and the default number of fields can be used to determine the default weighted value. Then, the product of the conflicting weighted data and the conflicting number of fields can be used to determine the conflicting weighted value. Finally, the sum of the explicitly weighted value, the default weighted value, and the conflicting weighted value can be used to determine the total field weighted value.

[0141] The third step involves performing weighted probability analysis on the number of explicit fields, the number of default fields, and the number of conflicting fields based on the total weight values ​​of the aforementioned fields, to obtain the ratios of explicit fields, default fields, and conflicting fields.

[0142] In practice, firstly, the ratio of the number of explicitly defined fields to the total weight of the fields can be determined as the explicitly defined field ratio. Secondly, the ratio of the number of default fields to the total weight of the fields can be determined as the default field ratio. Then, the ratio of the number of conflicting fields to the total weight of the fields can be determined as the conflicting field ratio.

[0143] The fourth step is to perform logarithmic weighting on the above-mentioned explicit field ratios, default field ratios, and conflict field ratios to obtain explicit weighted values, default weighted values, and conflict weighted values.

[0144] In practice, firstly, the natural logarithm of the aforementioned explicit field ratio can be determined as the explicit natural logarithm. Then, the product of the aforementioned explicit field ratio and the aforementioned explicit natural logarithm can be determined as the explicit weighted value.

[0145] Then, the natural logarithm of the above default field ratio can be determined as the default natural logarithm. Then, the product of the above default field ratio and the above default natural logarithm can be determined as the default weighted value.

[0146] Then, the natural logarithm of the aforementioned conflict field ratio can be determined as the conflict natural logarithm. Then, the product of the aforementioned conflict field ratio and the aforementioned conflict natural logarithm can be determined as the conflict weighting value.

[0147] The fifth step is to determine the sum of the explicitly weighted value, the default weighted value, and the conflicting weighted value as a logarithmically weighted sum.

[0148] Step 6: Based on the above logarithmic weighted sum, generate the field-weighted entropy. In practice, the negative of the above logarithmic weighted sum can be used as the field-weighted entropy.

[0149] Step 7: Determine the ratio of the above-mentioned field weighted entropy to the preset maximum weighted entropy as the field weighted entropy ratio. The above-mentioned maximum weighted entropy can be the natural logarithm of 3.

[0150] Step 8: Determine the drawing verification score by the difference between the preset value and the weighted entropy ratio of the above fields. The preset value can be 1.

[0151] Step 105: In response to determining that the drawing verification score is greater than the preset score threshold, the object processing equipment is controlled to perform the object processing task based on the drawing constraint database, the values ​​of each target field, each candidate field, and each conflict verification result.

[0152] In some embodiments, the execution entity may, in response to determining that the drawing verification score is greater than a preset score threshold, control the object processing equipment to perform an object processing task based on the drawing constraint database, the target fields, the candidate field values, and the conflict verification results.

[0153] In addressing the technical problems mentioned above, the application scenario—a precision machining workshop for automotive engine blocks—often presents the following challenges: when engineering drawings and global constraint documents conflict, relying solely on a single iteration of a large language model to determine the values ​​of conflicting fields can lead to low reliability of the determined values. This can result in machining errors based on these values, causing waste of materials and time. Considering the specific requirements of this application scenario—engine blocks often utilize high-strength cast iron or aluminum alloys, resulting in high material costs—waste of machining materials can lead to significant losses in the workshop. Furthermore, prolonged machining times can delay engine delivery. Therefore, to minimize waste of materials and time, we have decided to adopt the following solution:

[0154] In some optional implementations of certain embodiments, the execution entity can control the object processing equipment to perform object processing tasks based on the drawing constraint database, the target fields, the candidate field values, and the conflict verification results through the following steps:

[0155] For each of the preset names of objects to be processed, perform the following steps:

[0156] The first step involves filtering the target fields based on the names of the objects to be processed, resulting in object target fields. Each object name can be the name of the object to be processed. For example, the object name could be "white marble slab". Each object target field can be a target field related to the object name. In practice, the executing entity can identify the target fields that simultaneously contain both the object name and preset processing text as object target fields. The preset processing text can be a pre-defined string. For example, when the object name is "white marble slab" and the preset processing text is "size", the object target field could be "white marble slab size".

[0157] The second step, for each of the above object target fields, in response to determining that the conflict verification result corresponding to the above object target field meets the preset conflict conditions, is to perform the following steps:

[0158] The first sub-step involves converting the candidate field values ​​corresponding to the object target field into object processing information. The preset conflict condition can be either a conflict verification result of "no conflict" or "default fill value." The object processing information can be the instruction code corresponding to the candidate field values, which controls the object processing equipment to process the object.

[0159] In practice, the aforementioned executing entity can send the candidate field values ​​and preset processing instruction conversion information to the technical server. The processing instruction conversion information can be used to instruct the technical server to generate object processing information based on the candidate field values. For example, the processing instruction conversion information could be "Please convert the above candidate field values ​​into instructions that can control the object processing equipment to process the object." The technical server can be a server capable of generating object processing information based on the input candidate field values. Then, the entity can receive the data returned by the technical server as the object processing information.

[0160] The second sub-step involves controlling the object processing equipment to process the object corresponding to the name of the object to be processed, based on the aforementioned object processing information. In practice, the executing entity can send the object processing information to the object processing equipment to control the equipment to process the object corresponding to the name of the object to be processed.

[0161] Third, for each of the above object target fields, in response to determining that the conflict verification result corresponding to the above object target field does not meet the above preset conflict conditions, the following steps are performed:

[0162] The first sub-step involves generating conflict triples based on the candidate field values ​​corresponding to the aforementioned drawing constraint database and the aforementioned object target field. These conflict triples can be triples obtained from the aforementioned drawing constraint database and the aforementioned object target field.

[0163] In practice, firstly, the constraint fields stored in the drawing constraint database that are identical to the object target field can be identified as constraint fields to be processed. Then, the constraint field values ​​corresponding to these constraint fields can be identified as constraint field values ​​to be processed. Next, the candidate field values ​​corresponding to the object target field can be identified as the object target field values. Finally, the constraint fields to be processed, their values, and the object target field values ​​can be combined into triples as conflict triples.

[0164] For example, when the above constraint field to be processed is "marble slab length", the above constraint field value is "at least 25", and the above object target field value is "20", the conflict triple can be ("marble slab length", "at least 25", "20").

[0165] The second sub-step involves performing multi-path resolution processing on the aforementioned conflicting triples to obtain individual path-resolution triples. Each of these path-resolution triples can be a triple obtained after processing the conflicting triples.

[0166] In practice, firstly, the temperature parameters of the large language model can be adjusted to a preset temperature coefficient value using a parameter adjustment tool to update the large language model. This parameter adjustment tool can be an API capable of adjusting the parameters of the large language model. For example, when the large language model is GPT-4, the parameter adjustment tool used can be the OpenAI API. The preset temperature coefficient value can be a pre-defined value. For example, the preset temperature coefficient value can be 0.8. Secondly, the execution entity can perform the following processing steps: It can input the preset output prompt information and the aforementioned conflict triples into the updated large language model to obtain the data output by the large language model as path resolution triples. The aforementioned output prompt information can be information used to guide the large language model to output the corresponding format and content. For example, the output prompt information can be "Based on the first data in the conflict triple, determine which of the second and third data is the correct data, and output it in the following format (correct data, reason, whether manual review is still required)". Finally, the execution entity can perform the above processing steps a preset number of times to obtain each path resolution triple. The preset number can be a pre-defined value. For example, the preset quantity can be 20.

[0167] The third sub-step involves performing consistency aggregation on the aforementioned path resolution triples to obtain the target resolution triples. The target resolution triples can be the path resolution triples with the highest reliability.

[0168] In practice, firstly, the number of each path resolution triplet can be determined as the total number of triplets. For each path resolution triplet, the path resolution triplets with the same first value can be grouped into a set as a path resolution triplet set, thus obtaining each path resolution triplet set. Then, for each path resolution triplet set in the above sets, the number of each path resolution triplet in that set can be determined as the number of triplets. Then, the ratio of the number of triplets to the total number of triplets can be determined as the triplet aggregation degree. Then, the triplet aggregation degree with the highest value among the determined triplet aggregation degrees can be determined as the target triplet aggregation degree. Then, the path resolution triplet set corresponding to the target triplet aggregation degree can be determined as the target path resolution triplet set. Finally, any target path resolution triplet in the target path resolution triplet set can be determined as the target resolution triplet.

[0169] The fourth sub-step involves generating the object field value corresponding to the object target field based on the aforementioned target parsing triplet. In practice, the executing entity can determine the first data in the aforementioned target parsing triplet as the object field value.

[0170] The fifth sub-step involves controlling the object processing equipment to process the object corresponding to the name of the object to be processed, based on the aforementioned object field value.

[0171] In practice, the aforementioned object field values ​​and processing instruction conversion information can be sent to the aforementioned technical server. Then, the data returned by the technical server can be received as processing information. This processing information can then be sent to the object processing equipment to control the equipment to process the object corresponding to the name of the object to be processed.

[0172] The above-described technical solution and its related content, as an inventive point of this disclosure, solve the problem of "waste of processing materials". Factors leading to waste of processing materials and processing time often include: when there is a conflict between engineering drawings and global constraint documents, relying solely on a single iteration of the large language model to determine the field value corresponding to the conflicting field easily leads to low reliability of the determined field value, which in turn easily leads to processing errors when processing objects based on the field value, resulting in waste of processing materials and processing time. Solving these factors can reduce the waste of processing materials and time. To achieve this effect, this disclosure firstly, for each of the preset names of objects to be processed, performs the following steps: secondly, based on the aforementioned names of objects to be processed, performs data filtering processing on the aforementioned target fields to obtain the respective object target fields. Thus, the respective object target fields can be obtained. Then, for each of the aforementioned object target fields, in response to determining that the conflict verification result corresponding to the aforementioned object target field meets the preset conflict conditions, the following steps are performed: then, the candidate field value corresponding to the aforementioned object target field is converted into object processing information. Then, based on the aforementioned object processing information, the object processing equipment is controlled to process the object corresponding to the name of the object to be processed. Thus, when no conflict occurs, the object can be processed directly based on the object field value. Then, for each of the aforementioned object target fields, in response to determining that the conflict verification result corresponding to the aforementioned object target field does not meet the aforementioned preset conflict conditions, the following steps are executed: Then, based on the aforementioned drawing constraint database and the candidate field values ​​corresponding to the aforementioned object target fields, conflict triples are generated. Thus, when a conflict exists, conflict triples can be generated first. Then, multi-path resolution processing is performed on the aforementioned conflict triples to obtain each path resolution triplet. Thus, multi-path resolution processing can be performed on the conflict triples to obtain each path resolution triplet. Then, consistency aggregation processing is performed on the aforementioned path resolution triples to obtain the target resolution triplet. Thus, the target resolution triplet is obtained. Then, based on the aforementioned target resolution triplet, the object field value corresponding to the aforementioned object target field is generated. Thus, the object field value is obtained. Finally, based on the aforementioned object field value, the object processing equipment is controlled to process the object corresponding to the name of the object to be processed. Therefore, objects can be processed based on their field values. Furthermore, when there is a conflict between the candidate field values ​​corresponding to the drawing constraint database and the object's target field, the correct field value can be determined through multi-path parsing, rather than through a single iteration of the large language model. This improves the reliability of the generated candidate field values, thereby reducing the probability of processing errors due to low field value reliability and minimizing waste of processing materials and time.

[0173] The above embodiments of this disclosure have the following beneficial effects: The object processing method based on engineering drawing conflict verification according to some embodiments of this disclosure can reduce material waste during object processing. Specifically, the reason for material waste during object processing is that when data in the engineering drawing conflicts with the global description document, the large language model often has difficulty determining the correctness of the data, easily leading to low reliability of the data output by the large language model, and consequently, low reliability of the generated structured data. This results in processing errors when processing objects based on the structured data, leading to material waste. Based on this, the object processing method based on engineering drawing conflict verification according to some embodiments of this disclosure firstly, in response to receiving a processing signal sent by the object processing device, performs text extraction and constraint recognition processing on the pre-acquired global engineering drawing file to construct a drawing constraint database. Thus, constraint information can be extracted from the global engineering drawing file to construct the drawing constraint database. Secondly, based on preset target fields, data extraction and text association processing are performed on the pre-acquired local engineering drawing data to obtain candidate field values ​​and source annotation information. Thus, candidate field values ​​of the above-mentioned target fields can be extracted from the local engineering drawing data, and the source of the candidate field values ​​can be obtained. Then, for each of the aforementioned target fields, based on the source annotation information corresponding to the target field and the aforementioned drawing constraint database, data constraint matching and conflict verification processing are performed on the candidate field values ​​corresponding to the target field to obtain conflict verification results. Thus, the candidate field values ​​of each target field can be compared with the data in the drawing constraint database. When a candidate field value conflicts with the data in the drawing constraint database, the conflict verification result can be determined based on the source annotation information corresponding to the candidate field value. Then, the obtained conflict verification results are subjected to feasibility entropy quantification processing to obtain a drawing verification score. Thus, the feasibility of the engineering drawings can be evaluated based on the conflict verification results to obtain a drawing verification score. Finally, in response to determining that the drawing verification score is greater than a preset score threshold, the object processing equipment is controlled to execute the object processing task based on the aforementioned drawing constraint database, the aforementioned target fields, the aforementioned candidate field values, and the aforementioned conflict verification results. Therefore, when the reliability of the engineering drawings is high, the conflicting candidate field values ​​can be corrected using the conflict verification results and the drawing constraint database, and then the object processing task can be executed based on the corrected data.Because the candidate field values ​​of each target field can be extracted from the local data of the engineering drawings first, and the source annotation information of the candidate field values ​​can be marked, and when there is a conflict between the candidate field values ​​and the data in the drawing constraint database, the candidate field values ​​can be corrected according to the source annotation information and the conflict verification results, and then the object can be processed using the corrected data. Therefore, the reliability of each generated candidate field value can be improved, and the waste of processing materials caused by object processing errors can be reduced.

[0174] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of an object processing apparatus based on engineering drawing conflict verification. These apparatus embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0175] like Figure 2 As shown, an object processing apparatus 200 based on engineering drawing conflict verification in some embodiments includes: a construction unit 201, a first processing unit 202, a second processing unit 203, a quantization unit 204, and a control unit 205. The system comprises the following components: a construction unit 201, configured to, in response to receiving a processing signal from the object processing equipment, perform text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings to construct a drawing constraint database; a first processing unit 202, configured to, based on preset target fields, perform data extraction and text association processing on pre-acquired local data of engineering drawings to obtain candidate field values ​​and source annotation information; a second processing unit 203, configured to, for each of the target fields, perform data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field based on the source annotation information and the drawing constraint database to obtain conflict verification results; a quantization unit 204, configured to, perform feasibility entropy quantization processing on the obtained conflict verification results to obtain a drawing verification score; and a control unit 205, configured to, in response to determining that the drawing verification score is greater than a preset score threshold, control the object processing equipment to execute an object processing task based on the drawing constraint database, the target fields, the candidate field values, and the conflict verification results.

[0176] It is understandable that the units described in the device 200 are related to the reference. Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 200 and the units contained therein, and will not be repeated here.

[0177] The following is for reference. Figure 3It shows a schematic diagram of the structure of an electronic device (such as a computing device) 300 suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0178] like Figure 3 As shown, the electronic device 300 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0179] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0180] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0181] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0182] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0183] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: in response to receiving a processing signal sent by the object processing device, perform text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings to construct a drawing constraint database; based on preset target fields, perform data extraction and text association processing on the pre-acquired local data of the engineering drawings to obtain candidate field values ​​and source annotation information; for each of the aforementioned target fields, based on the source annotation information corresponding to the target field and the aforementioned drawing constraint database, perform data constraint matching and conflict verification processing on the candidate field values ​​corresponding to the target field to obtain conflict verification results; perform feasibility entropy quantification processing on the obtained conflict verification results to obtain a drawing verification score; in response to determining that the drawing verification score is greater than a preset score threshold, control the aforementioned object processing device to execute an object processing task based on the aforementioned drawing constraint database, the aforementioned target fields, the aforementioned candidate field values, and the aforementioned conflict verification results.

[0184] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0185] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0186] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including a building unit, a first processing unit, a second processing unit, a quantization unit, and a control unit. The names of these units do not necessarily limit the specific unit; for example, a control unit may also be described as "a unit that controls the aforementioned object processing equipment to perform object processing tasks."

[0187] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0188] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. An object processing method based on engineering drawing conflict checking, characterized in that, include: In response to receiving a processing signal from an object processing device, text extraction and constraint recognition processing are performed on a pre-acquired global file of engineering drawings to construct a drawing constraint database. The method is characterized in that the text extraction and constraint recognition processing on the pre-acquired global file of engineering drawings to construct the drawing constraint database includes: The global file of the engineering drawing is processed by drawing frame recognition to obtain global drawing frame data. Each global drawing frame data corresponds to a drawing frame name and boundary coordinate data. Based on the names of the corresponding global frame data, data filtering is performed on the global frame data to obtain the target frame data. Based on the boundary coordinate data corresponding to each target frame data, the text content of the global file of the engineering drawing is extracted to obtain the text data of each image. Based on each preset target string, text filtering processing is performed on each image text data to obtain each target image text data; The constraint condition recognition processing is performed on the text data of each target image to obtain constrained structured data. The process is characterized in that the constraint condition recognition processing of the text data of each target image to obtain constrained structured data includes: For each target image text data in the aforementioned target image text data, perform the following steps: Based on the preset field data table and the frame name corresponding to the target image text data, determine each image field corresponding to the target image text data; Based on the image fields, the target image text data is subjected to field value matching processing to obtain image key-value pair data, wherein each image key-value pair data includes an image field and an image field value; The image key-value pair data is subjected to structured cleaning processing to obtain cleaned key-value pair data. The data for each cleaning key value is normalized to obtain image text structured data; Logical conflict verification and conflict resolution are performed on the image text structured data to obtain constrained structured data; Based on the aforementioned structured constraint data, a drawing constraint database is constructed; Based on the preset target fields, data extraction and text association processing are performed on the pre-acquired partial data of the engineering drawings to obtain the values ​​of each candidate field and the annotation information of each source; For each of the target fields, based on the source annotation information corresponding to the target field and the drawing constraint database, data constraint matching and conflict verification processing are performed on the candidate field values ​​corresponding to the target field to obtain the conflict verification result; The feasibility entropy values ​​of each conflict verification result are quantified to obtain the drawing verification score; In response to determining that the drawing verification score is greater than a preset score threshold, the object processing equipment is controlled to perform an object processing task based on the drawing constraint database, the target fields, the candidate field values, and the conflict verification results.

2. The method of claim 1, wherein, The drawing constraint database stores structured data for each constraint, and each structured constraint data includes a constraint field value; and based on the source annotation information corresponding to the target field and the drawing constraint database, data constraint matching and conflict verification processing are performed on the candidate field values ​​corresponding to the target field to obtain conflict verification results, including: The constraint field value corresponding to the target field is retrieved from the drawing constraint database and used as the target constraint field value; In response to determining that the source annotation information corresponding to the target field meets the preset missing annotation conditions, based on the target constraint field value, data filling is performed on the candidate field value corresponding to the target field to update the candidate field value, and the preset field conflict information is determined as the conflict verification result; In response to determining that the source annotation information corresponding to the target field meets the preset explicit annotation conditions, the following steps are performed: In response to determining that the candidate field value meets the preset text data conditions, conflict detection processing is performed on the candidate field value and the target constraint field value to obtain a conflict verification result; In response to determining that the candidate field value meets the preset numerical data conditions, the following steps are performed: The target constraint field values ​​are logically standardized to obtain logical constraint objects; Based on the logical constraint object, unit conversion and conflict detection processing are performed on the candidate field values ​​corresponding to the target field to obtain the conflict detection result. Based on the conflict detection results, a conflict verification result is generated.

3. The method according to claim 2, characterized in that, in After generating the conflict verification result based on the conflict detection result, the method further includes: In response to determining that the conflict verification result satisfies the preset conflict verification conditions, the following steps are performed: Based on the preset first prompt information and the preset second prompt information, multi-agent conflict resolution is performed on the target constraint field value and the candidate field value corresponding to the target field to obtain the first conflict resolution information and the second conflict resolution information; Based on the preset conflict prompt information, the target constraint field value, and the candidate field value corresponding to the target field, the first conflict resolution information and the second conflict resolution information are intelligently parsed to obtain the third conflict resolution information; Based on the third conflict resolution information, the target constraint field value, and the candidate field value corresponding to the target field, an adjustment field value corresponding to the target field is generated; Based on the target field, the value of the field to be adjusted, and the preset adjustment information, the value of the field to be adjusted is adaptively adjusted to obtain the target field value; Combine the target field and the target field value into target key-value pair data.

4. The method according to claim 1, characterized in that, The feasibility entropy value quantification process is performed on each of the obtained conflict verification results to obtain the drawing verification score, including: Based on the conflict verification results, the number of explicit fields, the number of default fields, and the number of conflicting fields are determined. Based on preset explicit weight data, preset default weight data, and preset conflict weight data, the number of explicit fields, the number of default fields, and the number of conflict fields are weighted to obtain the total field weight value. Based on the total weight value of the fields, the number of explicit fields, the number of default fields, and the number of conflicting fields are analyzed by weight probability to obtain the ratio of explicit fields, the ratio of default fields, and the ratio of conflicting fields; Logarithmically weighted the explicit field ratio, the default field ratio, and the conflicting field ratio to obtain explicit weighted values, default weighted values, and conflicting weighted values; The sum of the explicitly weighted value, the default weighted value, and the conflicting weighted value is determined as a logarithmic weighted sum; Based on the logarithmic weighted sum, generate the field weighted entropy; The ratio of the field weighted entropy to the preset maximum weighted entropy is determined as the field weighted entropy ratio; The difference between the preset value and the weighted entropy ratio of the field is determined as the drawing verification score.

5. An object processing device based on conflict verification of engineering drawings, characterized in that, include: A construction unit, configured to perform text extraction and constraint recognition processing on a pre-acquired global file of engineering drawings in response to receiving a processing signal from an object processing device, to construct a drawing constraint database, characterized in that the text extraction and constraint recognition processing on the pre-acquired global file of engineering drawings to construct the drawing constraint database includes: The global file of the engineering drawing is processed by drawing frame recognition to obtain global drawing frame data. Each global drawing frame data corresponds to a drawing frame name and boundary coordinate data. Based on the names of the corresponding global frame data, data filtering is performed on the global frame data to obtain the target frame data. Based on the boundary coordinate data corresponding to each target frame data, the text content of the global file of the engineering drawing is extracted to obtain the text data of each image. Based on each preset target string, text filtering processing is performed on each image text data to obtain each target image text data; The constraint condition recognition processing is performed on the text data of each target image to obtain constrained structured data. The process is characterized in that the constraint condition recognition processing of the text data of each target image to obtain constrained structured data includes: For each target image text data in the aforementioned target image text data, perform the following steps: Based on the preset field data table and the frame name corresponding to the target image text data, determine each image field corresponding to the target image text data; Based on the image fields, the target image text data is subjected to field value matching processing to obtain image key-value pair data, wherein each image key-value pair data includes an image field and an image field value; The image key-value pair data is subjected to structured cleaning processing to obtain cleaned key-value pair data. The data for each cleaning key value is normalized to obtain image text structured data; Logical conflict verification and conflict resolution are performed on the image text structured data to obtain constrained structured data; Based on the aforementioned structured constraint data, a drawing constraint database is constructed; The first processing unit is configured to perform data extraction and text association processing on the pre-acquired partial data of the engineering drawings based on preset target fields, so as to obtain the values ​​of each candidate field and the annotation information of each source. The second processing unit is configured to perform data constraint matching and conflict verification processing on the candidate field value corresponding to the target field for each of the target fields, based on the source annotation information corresponding to the target field and the drawing constraint database, and obtain the conflict verification result. The quantization unit is configured to quantify the feasibility entropy value of each conflict verification result to obtain the drawing verification score; The control unit is configured to, in response to determining that the drawing verification score is greater than a preset score threshold, control the object processing equipment to perform an object processing task based on the drawing constraint database, the respective target fields, the respective candidate field values, and the respective conflict verification results.

6. An electronic device, characterized in that, include: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 4.

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

Images