Search

CN-122024266-A - Engineering contract identification method and system based on image identification

CN122024266ACN 122024266 ACN122024266 ACN 122024266ACN-122024266-A

Abstract

The application relates to an engineering contract identification method and system based on image identification, and relates to the field of image processing, wherein the engineering contract identification method and system comprises the steps of sequentially carrying out denoising treatment, binarization treatment and rotation correction on a combined image sequence to obtain a standard contract image sequence; the method comprises the steps of obtaining a reference contract image sequence, executing pattern alignment, pattern segmentation and mapping combination of the reference contract image sequence and a standard contract image sequence to obtain a local image sequence group, inputting the local image sequence group into a contract identification model for deviation analysis, outputting a contract deviation information set, carrying out content identification according to the contract deviation information set, and outputting a contract content identification result. The application can solve the technical problems of high contract recognition cost and low recognition efficiency caused by the insufficient universality of the recognition model in the existing method, can improve the intelligentization degree of contract recognition, realizes efficient and low-cost recognition and content analysis of engineering contracts, and remarkably improves the recognition efficiency and accuracy of the engineering contracts.

Inventors

  • HUANG DAQUAN

Assignees

  • 华腾建信科技有限公司

Dates

Publication Date
20260512
Application Date
20260326

Claims (10)

  1. 1. An engineering contract identification method based on image identification is characterized by comprising the following steps: acquiring images of engineering contracts to be identified by using a CCD image sensor to acquire contract image sequences; according to a preset processing scheme, denoising, binarization and rotation correction are sequentially carried out on the contract image sequence to obtain a standard contract image sequence; acquiring a reference contract image sequence of the engineering contract to be identified, and executing pattern alignment, pattern segmentation and mapping combination of the reference contract image sequence and the standard contract image sequence to obtain a local image sequence group; Inputting the local image sequence group into a contract identification model for deviation analysis, and outputting a contract deviation information set, wherein the contract identification model is constructed based on a twin neural network; and carrying out content identification according to the contract deviation information set, and outputting a contract content identification result.
  2. 2. The method for recognizing engineering contracts based on image recognition according to claim 1, wherein the step of sequentially performing denoising, binarization and rotation correction on the contract image sequence to obtain a standard contract image sequence comprises: removing noise points of the contract image sequence by using a Gaussian filtering algorithm, and carrying out contrast enhancement by using a histogram equalization algorithm to obtain a primary processing image sequence; and carrying out binarization processing and image rotation correction on the once processed image sequence, and outputting the standard contract image sequence.
  3. 3. The method for recognizing engineering contracts based on image recognition according to claim 1, wherein executing the pattern alignment, pattern segmentation and mapping combination of the reference contract image sequence and the standard contract image sequence to obtain the partial image sequence group includes: configuring a layout segmentation rule based on a layout structure of the reference contract image sequence; Executing the layout segmentation of the aligned contract images according to the layout segmentation rules, and determining a local reference image sequence and a local standard image sequence; and carrying out one-to-one mapping combination on the local reference image and the local standard image to obtain the local image sequence group.
  4. 4. The method for recognizing engineering contracts based on image recognition according to claim 1, wherein inputting the partial image sequence group into a contract recognition model for deviation analysis, outputting a contract deviation information set, comprises: constructing a contract identification model based on a twin neural network, wherein the contract identification model comprises a feature extraction layer and a deviation comparison layer; Inputting the local image sequence group into the contract identification model, and extracting character features by utilizing the feature extraction layer to obtain a plurality of character sequence groups; And carrying out deviation analysis on the plurality of character sequence groups through the deviation comparison layer, and outputting the contract deviation information set.
  5. 5. The method for identifying engineering contracts based on image identification according to claim 4, wherein constructing the feature extraction layer includes: constructing a feature extraction layer, wherein the feature extraction layer comprises two convolution sub-networks with shared weights; And collecting a sample contract image set and a sample character data set as training data, performing supervision training on the feature extraction layer until convergence, and outputting the feature extraction layer after training is completed.
  6. 6. The method for recognizing engineering contracts based on image recognition according to claim 4, wherein the performing of the deviation analysis on the plurality of character sequence groups by the deviation comparison layer includes: Constructing a deviation comparison layer, wherein the deviation comparison layer comprises a similarity comparison unit and a deviation judgment unit, and the deviation judgment unit is embedded with a preset similarity threshold; inputting the plurality of character sequences into the similarity comparison unit respectively, and outputting a plurality of similarity sets; And judging the plurality of similarity sets according to the preset similarity threshold value, and selecting character features smaller than the preset similarity threshold value as contract deviation information.
  7. 7. The method for recognizing engineering contracts based on image recognition according to claim 6, wherein the construction of the similarity comparison unit includes: Configuring a similarity comparison operator, wherein the similarity comparison operator at least comprises a cosine comparison algorithm, a Euclidean distance and a Manhattan distance; constructing a first comparison branch, a second comparison branch and a third comparison branch according to the cosine comparison algorithm, the Euclidean distance and the Manhattan distance; And integrating and constructing the similar comparison unit based on the first comparison branch, the second comparison branch and the third comparison branch, wherein the output of the similar comparison unit is the average value of the output results of the first comparison branch, the second comparison branch and the third comparison branch.
  8. 8. An image recognition-based engineering contract recognition system, characterized by the steps for implementing an image recognition-based engineering contract recognition method according to any one of claims 1 to 7, comprising: The contract image sequence acquisition module is used for acquiring images of engineering contracts to be identified by using the CCD image sensor to acquire contract image sequences; The contract image processing module is used for sequentially carrying out denoising processing, binarization processing and rotation correction on the contract image sequence according to a preset processing scheme to obtain a standard contract image sequence; the local image sequence group obtaining module is used for obtaining a reference contract image sequence of the engineering contract to be identified, executing pattern alignment, pattern segmentation and mapping combination of the reference contract image sequence and the standard contract image sequence, and obtaining a local image sequence group; the contract deviation analysis module is used for inputting the local image sequence group into a contract identification model for deviation analysis and outputting a contract deviation information set, wherein the contract identification model is constructed based on a twin neural network; And the contract content identification module is used for carrying out content identification according to the contract deviation information set and outputting a contract content identification result.
  9. 9. An electronic device, comprising: A memory for storing a computer software program; A processor for reading and executing the computer software program to implement the steps of an image recognition-based engineering contract recognition method according to any one of claims 1 to 7.
  10. 10. A non-transitory computer readable storage medium, characterized in that the storage medium has stored therein a computer software program which, when executed by a processor, implements the steps of an image recognition based engineering contract recognition method according to any one of claims 1 to 7.

Description

Engineering contract identification method and system based on image identification Technical Field The invention relates to the field of image processing, in particular to an engineering contract identification method and system based on image identification. Background The existing engineering contract identification method generally needs to construct a special identification model for each contract category and symbol, has the problems of poor model reusability and difficult migration, and needs to redesign and train the model when facing complex scenes or newly added contract categories, so that the model construction and maintenance cost is high. Therefore, a new method is needed that enables efficient identification of diverse contracts without adding additional model building costs. Disclosure of Invention The invention provides an engineering contract identification method and an engineering contract identification system based on image identification, aiming at the technical problems of higher contract identification cost and lower identification efficiency caused by the fact that the existing engineering contract identification method based on image identification has insufficient universality of an identification model. The technical scheme for solving the technical problems is as follows: The invention provides an engineering contract identification method based on image identification, which comprises the steps of utilizing a CCD image sensor to acquire an image of an engineering contract to be identified, obtaining a contract image sequence, sequentially carrying out denoising, binarization and rotation correction on the contract image sequence according to a preset processing scheme to obtain a standard contract image sequence, obtaining a standard contract image sequence of the engineering contract to be identified, executing pattern alignment, pattern segmentation and mapping combination of the standard contract image sequence and the standard contract image sequence to obtain a local image sequence group, inputting the local image sequence group into a contract identification model for deviation analysis, and outputting a contract deviation information set, wherein the contract identification model is constructed based on a twin neural network, carrying out content identification according to the contract deviation information set, and outputting a contract content identification result. The invention provides an engineering contract identification system based on image identification, which comprises a contract image sequence acquisition module, a contract image processing module, a local image sequence group acquisition module, a contract content identification module and a contract deviation analysis module, wherein the contract image sequence acquisition module is used for acquiring an image of an engineering contract to be identified by using a CCD (charge coupled device) image sensor to acquire a contract image sequence, the contract image processing module is used for sequentially carrying out denoising processing, binarization processing and rotation correction on the contract image sequence according to a preset processing scheme to acquire a standard contract image sequence, the local image sequence group acquisition module is used for acquiring a reference contract image sequence of the engineering contract to be identified, and executing plate alignment, plate segmentation and mapping combination of the reference contract image sequence and the standard contract image sequence to acquire a local image sequence group, the contract deviation analysis module is used for inputting the local image sequence group into a contract identification model to carry out deviation analysis to output a contract deviation information set, wherein the contract identification model is constructed based on a twin neural network, and the contract content identification module is used for carrying out content identification according to the contract deviation information set to output contract content identification result. In a third aspect, the present invention also provides an electronic device, including: The system comprises at least one processor, a memory communicatively coupled to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the steps of the method of any one of the first aspects. In a fourth aspect, a computer readable storage medium having stored thereon a computer program which, when executed, implements the steps of the method of any of the first aspects above. The method has the advantages that the CCD image sensor is utilized to acquire images of engineering contracts to be identified to obtain a contract image sequence, denoising, binarization and rotation correction are sequentially carried out on the contract image sequence according to a preset processing scheme to obtai