Search

CN-122024058-A - Image recognition-based light inner partition plate defect detection method and system

CN122024058ACN 122024058 ACN122024058 ACN 122024058ACN-122024058-A

Abstract

The application discloses a method and a system for detecting defects of a lightweight inner partition plate based on image recognition, and relates to the technical field of image recognition, wherein the method comprises the steps of collecting images of the surface of the partition plate for surface crack detection and obtaining surface macrocrack state distribution; the method comprises the steps of predicting the distribution of the micro-cracks on the surface of the wallboard, obtaining the position distribution of the micro-cracks on the surface of the wallboard, determining the distribution of the micro-crack areas on the surface, optimizing the irradiation parameters of a detection light source, determining the irradiation parameters of an optimal light source, taking the distribution of the micro-crack areas on the surface as a region to be detected, irradiating the detection light source on the micro-crack areas on the surface, collecting and obtaining a second wallboard surface image set, detecting the micro-cracks on the surface to obtain the state distribution of the micro-cracks on the surface, and generating the wallboard surface crack detection result. The technical problems that the existing manual visual detection intensity is high, the efficiency is low, the result is easily influenced by subjective factors, and the defect detection accuracy and reliability are low are solved.

Inventors

  • LI LINGYE

Assignees

  • 广西科技大学

Dates

Publication Date
20260512
Application Date
20260130

Claims (10)

  1. 1. The method for detecting the defects of the light inner partition plate based on the image recognition is characterized by comprising the following steps of: acquiring wallboard surface images of the light inner partition boards under the daily light condition to detect surface cracks and acquire surface macrocrack state distribution; Predicting the micro-crack distribution on the surface of the wallboard based on the surface micro-crack state distribution, obtaining predicted surface micro-crack position distribution, and performing region self-adaptive expansion to determine surface micro-crack region distribution; Optimizing the irradiation parameters of the detection light source by taking the contrast ratio of the increased microcracks and the wallboard background as an optimizing target, and determining the optimal light source irradiation parameters; Taking the surface micro-crack area distribution as an area to be detected, carrying out detection light source irradiation on the surface micro-crack area according to the optimal light source irradiation parameters, and acquiring a second wallboard surface image set; and carrying out surface micro-crack detection according to the second wallboard surface image set to obtain surface micro-crack state distribution, and generating wallboard surface crack detection results by combining the surface micro-crack state distribution.
  2. 2. The image recognition-based defect detection method for the light inner partition plate, which is disclosed in claim 1, is characterized in that surface crack detection is carried out by collecting the wallboard surface images of the light inner partition plate under the daily light condition, and the surface macro-crack state distribution is obtained, wherein the light inner partition plate is an autoclaved light concrete plate, and the crack state comprises crack type, crack trend and crack size.
  3. 3. The image recognition-based light-weight interior partition plate defect detection method according to claim 1, wherein the performing of the wall plate surface micro-crack distribution prediction based on the surface micro-crack state distribution to obtain the predicted surface micro-crack position distribution comprises: Calculating to obtain the surface density of the macrocracks and the number density of the macrocracks based on the surface macrocrack state distribution; Performing crack distribution discreteness evaluation based on the surface macrocrack state distribution to obtain macrocrack distribution discreteness; determining a micro-crack prediction complexity index according to the macro-crack surface density, the macro-crack number density and the macro-crack distribution dispersion degree evaluation, wherein the micro-crack prediction complexity index is positively correlated with the macro-crack surface density and the macro-crack number density and negatively correlated with the macro-crack distribution dispersion degree; and calling a micro-crack prediction channel according to the micro-crack prediction complex index, predicting the micro-crack distribution of the wallboard surface according to the surface macro-crack state distribution, and outputting predicted surface micro-crack position distribution.
  4. 4. The image recognition-based light-weight inner partition plate defect detection method of claim 3, wherein invoking a micro-crack prediction channel according to the micro-crack prediction complexity index, performing wall plate surface micro-crack distribution prediction according to the surface macro-crack state distribution, and outputting predicted surface micro-crack position distribution, comprising: A pre-training micro-crack prediction channel, wherein the micro-crack prediction channel comprises K micro-crack prediction branches, K is an integer greater than 20; Multiplying the ratio of the micro-crack prediction complex index to the preset micro-crack prediction complex index by the initial micro-crack prediction branch selection number L and rounding up to obtain the optimal micro-crack prediction branch selection number H, wherein L is 6, H is equal to 3 if the calculated H is smaller than 3, and H is equal to K if the calculated H is larger than K; randomly selecting H predicted branches from the K micro-crack predicted branches, respectively predicting according to the surface macro-crack state distribution, and outputting H initial predicted surface micro-crack position distribution; And counting the union of the surface micro-crack positions in the H initial predicted surface micro-crack position distributions as the predicted surface micro-crack position distribution.
  5. 5. The method for detecting defects of a lightweight interior partition panel based on image recognition according to claim 4, wherein the pre-training of the microcrack prediction channel comprises: taking wallboard attribute characteristics of the light inner partition boards as constraints, carrying out positive sample retrieval based on industrial big data, collecting sample surface macrocrack state distribution sets, and collecting historical surface microcrack position distributions corresponding to different sample surface macrocrack state distributions as training labels to obtain a training label set; taking the sample surface macrocrack state distribution set and the training label set as training data, and carrying out K-fold cross division with back placement to obtain K sample training sets; And training a deep learning model to converge by using the K sample training sets by using the sample surface macrocrack state distribution as input and the training labels as supervision to obtain K microcrack prediction branches.
  6. 6. The image recognition-based light-weight inner partition plate defect detection method according to claim 1, wherein the determining of the surface microcrack area distribution by performing area adaptive expansion comprises the following steps: Marking the surface macrocrack positions based on the surface macrocrack state distribution to obtain the surface macrocrack position distribution; performing position mapping on the surface macrocrack position distribution and the predicted surface microcrack position distribution, calculating the shortest distance between each predicted surface microcrack position and the nearest surface macrocrack position, and setting the area expansion size according to the shortest distance to obtain area expansion size distribution; and carrying out area self-adaptive expansion on the predicted surface micro-crack position distribution according to the area expansion size distribution to generate surface micro-crack area distribution.
  7. 7. The image recognition-based light-weight inner partition plate defect detection method according to claim 6, wherein the area expansion size is obtained by multiplying the ratio of the shortest distance to a preset standard interval distance by an initial area expansion size.
  8. 8. The image recognition-based light-weight inner partition plate defect detection method according to claim 1, wherein the optimizing of the irradiation parameters of the detection light source by taking the contrast of the microcrack and the wall plate background as an optimizing target, and the determining of the optimal light source irradiation parameters comprises the following steps: acquiring an irradiation parameter adjustment space of a detection light source, wherein the irradiation parameters comprise a center wavelength, an irradiation angle and illuminance; Randomly selecting a plurality of initial irradiation parameters in the irradiation parameter adjustment space, and respectively combining the initial irradiation parameters with wallboard attribute characteristics of the light inner partition board to obtain a plurality of light source irradiation schemes; respectively carrying out irradiation simulation according to the light source irradiation schemes in the light source irradiation simulation space, and taking the gray level difference of the microcracks and the wallboard background as simulation contrast to obtain a plurality of simulation contrast; And optimizing the irradiation parameters of the detection light source according to the initial irradiation parameters and the simulated contrasts, and determining the optimal light source irradiation parameters.
  9. 9. The method for detecting defects of light-weight inner partition boards based on image recognition according to claim 8, wherein the optimizing of the irradiation parameters of the detection light source according to the initial irradiation parameters and the simulated contrasts comprises: Setting an initial irradiation parameter as an initial solution, arranging a plurality of initial solutions from large to small according to the simulated contrast ratio to generate an initial solution sequence, taking the first solution of the initial solution sequence as an optimal solution, and taking the remaining solution as an inferior solution; taking the optimal solution as an adjustment direction, adjusting a plurality of inferior solutions according to a preset optimizing step length, and generating a plurality of updated inferior solutions; Reordering the optimal solution and the plurality of updated inferior solutions according to the sequence of the simulation contrast from large to small, eliminating the terminal inferior solutions with preset proportion, and performing equivalent supplement in the irradiation parameter adjustment space to generate an updated solution sequence, wherein the preset proportion gradually decreases along with the increase of optimizing times; and according to an iterative optimizing mechanism of optimal solution selection, inferior demodulation and adjustment, inferior solution elimination and initial solution supplementation, continuing to optimize the irradiation parameters according to the updated solution sequence until the preset convergence times are reached, and outputting the optimal solution of the current updated solution sequence as the optimal light source irradiation parameters.
  10. 10. A light interior partition panel defect detection system based on image recognition, for performing the light interior partition panel defect detection method based on image recognition of any one of claims 1-9, comprising: the data acquisition module is used for acquiring wallboard surface images of the light inner partition boards under the daily light condition to detect surface cracks and acquire surface macrocrack state distribution; The distribution prediction module is used for predicting the micro-crack distribution on the surface of the wallboard based on the surface micro-crack state distribution, obtaining the predicted surface micro-crack position distribution, and performing region self-adaptive expansion to determine the surface micro-crack region distribution; the parameter optimization module is used for optimizing the irradiation parameters of the detection light source by taking the contrast ratio of the increased microcracks and the wallboard background as an optimizing target, and determining the optimal irradiation parameters of the light source; The detection execution module is used for taking the surface micro-crack area distribution as an area to be detected, carrying out detection light source irradiation on the surface micro-crack area according to the optimal light source irradiation parameters, and acquiring a second wallboard surface image set; and the result generation module is used for carrying out surface micro-crack detection according to the second wallboard surface image set to obtain surface micro-crack state distribution, and generating wallboard surface crack detection results by combining the surface micro-crack state distribution.

Description

Image recognition-based light inner partition plate defect detection method and system Technical Field The application relates to the technical field of image recognition, in particular to a method and a system for detecting defects of a light inner partition plate based on image recognition. Background Along with the rapid development of building industrialization, the light inner partition board is widely applied to non-bearing walls of various buildings due to the advantages of light weight, high strength, environmental protection, convenient construction and the like, but cracks are easily generated on the surface of the light inner partition board, so that the appearance quality of the wall board is influenced, and the structural performance and durability of the wall board are weakened. Meanwhile, the autoclaved lightweight concrete slab has the advantages of high macroscopic flatness, low microscopic roughness, high homogeneity of materials and uniform color, so that shadows are hardly generated under the conventional uniform light, the gray level difference between the autoclaved lightweight concrete slab and the background is extremely small, and the surface microscopic cracks are difficult to identify. The defect detection of the existing light inner partition plate depends on manual visual inspection, so that the intensity is high, the efficiency is low, the detection result is easily influenced by subjective factors of detection staff, the objectivity and consistency of detection are difficult to ensure, and the omission ratio is high and the accuracy and reliability are low. Disclosure of Invention The embodiment of the application solves the technical problems of high strength, low efficiency and low defect detection accuracy and reliability caused by the fact that the conventional manual visual detection is easily influenced by subjective factors by providing the method and the system for detecting the defects of the light inner partition plate based on image recognition. The technical scheme for solving the technical problems is as follows: In a first aspect, the present application provides a method for detecting defects of a lightweight inner partition board based on image recognition, the method comprising: acquiring wallboard surface images of the light inner partition boards under the daily light condition to detect surface cracks and acquire surface macrocrack state distribution; Predicting the micro-crack distribution on the surface of the wallboard based on the surface micro-crack state distribution, obtaining predicted surface micro-crack position distribution, and performing region self-adaptive expansion to determine surface micro-crack region distribution; Optimizing the irradiation parameters of the detection light source by taking the contrast ratio of the increased microcracks and the wallboard background as an optimizing target, and determining the optimal light source irradiation parameters; Taking the surface micro-crack area distribution as an area to be detected, carrying out detection light source irradiation on the surface micro-crack area according to the optimal light source irradiation parameters, and acquiring a second wallboard surface image set; and carrying out surface micro-crack detection according to the second wallboard surface image set to obtain surface micro-crack state distribution, and generating wallboard surface crack detection results by combining the surface micro-crack state distribution. In a second aspect, the present application provides an image recognition-based defect detection system for a lightweight interior partition panel, comprising: the data acquisition module is used for acquiring wallboard surface images of the light inner partition boards under the daily light condition to detect surface cracks and acquire surface macrocrack state distribution; The distribution prediction module is used for predicting the micro-crack distribution on the surface of the wallboard based on the surface micro-crack state distribution, obtaining the predicted surface micro-crack position distribution, and performing region self-adaptive expansion to determine the surface micro-crack region distribution; the parameter optimization module is used for optimizing the irradiation parameters of the detection light source by taking the contrast ratio of the increased microcracks and the wallboard background as an optimizing target, and determining the optimal irradiation parameters of the light source; The detection execution module is used for taking the surface micro-crack area distribution as an area to be detected, carrying out detection light source irradiation on the surface micro-crack area according to the optimal light source irradiation parameters, and acquiring a second wallboard surface image set; and the result generation module is used for carrying out surface micro-crack detection according to the second wallboard surface image set to obtain surfac