Search

CN-121981393-A - House source intelligent house checking method and system based on computer vision and multi-sensor fusion

CN121981393ACN 121981393 ACN121981393 ACN 121981393ACN-121981393-A

Abstract

The application provides a house source intelligent house checking method and system based on computer vision and multi-sensor fusion, and belongs to the field of intelligent house checking. The method comprises the steps of obtaining multi-mode sensing data such as visual images, three-dimensional point clouds, environmental parameters and equipment operation, carrying out space-time registration and feature fusion to construct a multi-dimensional room inspection feature set, combining defect identification and parameter deviation analysis, establishing a space quality coupling model, mining cross-mode relevance to generate a defect risk mapping network, introducing a dynamic defect prediction model to conduct prediction evaluation on defect evolution trend and degradation degree to form a room source quality risk map, establishing a multi-dimensional quality evaluation system to realize comprehensive grade assessment of the risk map, outputting a structural laboratory report, generating a rectification decision suggestion based on the report, and realizing on-site real-time early warning and scheme pushing by combining edge calculation, so that the automation, the intelligence and the forward prospectivity of the room inspection process are realized, and the detection precision, the risk pre-judging capability and the management efficiency are improved.

Inventors

  • DING JUN
  • XIAO DONG

Assignees

  • 上海蝉觉网络科技有限公司

Dates

Publication Date
20260505
Application Date
20260126

Claims (10)

  1. 1. A house source intelligent house checking method based on computer vision and multi-sensor fusion, which is characterized by comprising the following steps: acquiring multi-mode sensing data of a room source space, and performing space-time registration and feature fusion on the multi-mode sensing data to acquire a multi-dimensional room inspection feature set; performing defect identification and parameter deviation analysis on the multidimensional room inspection feature set, constructing a space quality coupling model, and analyzing cross-modal feature relevance to obtain a defect risk mapping network; constructing a dynamic defect prediction model, and carrying out evolution trend prediction and severity assessment on a defect risk mapping network based on the dynamic defect prediction model so as to obtain a house source quality risk map; establishing a house quality evaluation index system, and carrying out comprehensive grade evaluation on a house quality risk map to obtain a house verification report; Generating a correction decision proposal according to the house source house inspection report, and realizing real-time defect early warning and correction proposal pushing through edge calculation.
  2. 2. The room source intelligent room inspection method based on computer vision and multi-sensor fusion of claim 1, wherein the acquiring multi-modal sensing data of room source space, performing space-time registration and feature fusion on the multi-modal sensing data, and obtaining a multi-dimensional room inspection feature set, comprises: acquiring multi-mode sensing data of a room source space, wherein the multi-mode sensing data comprises visual image data, three-dimensional point cloud data, environment sensing data and equipment operation data; performing defect feature extraction on the visual image data to obtain surface defect features; Performing space coordinate conversion and size verification on the three-dimensional point cloud data to obtain structural deformation characteristics; Performing time domain fluctuation analysis on the environmental sensing data to obtain environmental quality characteristics; Extracting parameter trend from the equipment operation data to obtain functional performance characteristics; performing space coupling analysis based on the surface defect characteristics and the structural deformation characteristics to obtain a physical defect association coefficient; Correcting the environmental quality characteristics based on the physical defect association coefficient to obtain real environmental influence characteristics; and carrying out space-time registration based on the real environment influence features, the surface defect features, the structural deformation features and the functional performance features to construct a multi-dimensional room inspection feature set.
  3. 3. The room source intelligent checking method based on the integration of computer vision and multiple sensors as claimed in claim 2, wherein the performing defect feature extraction on the visual image data to obtain surface defect features comprises: Performing pixel-level defect identification on the image by adopting a deep learning semantic segmentation model to obtain defect region mask data; calculating the area, perimeter, maximum width and deviation amount of the defect area and the standard threshold value, and constructing a defect parameter matrix; performing morphological analysis based on the defect parameter matrix to generate a defect type classification result; and calculating defect density and relative severity of different areas on the same surface, and constructing defect feature vectors to form surface defect features.
  4. 4. The room source intelligent room inspection method based on the integration of computer vision and multiple sensors according to claim 1, wherein the steps of performing defect recognition and parameter deviation analysis on the multi-dimensional room inspection feature set, constructing a spatial quality coupling model, analyzing cross-modal feature relevance and obtaining a defect risk mapping network include: the surface defects, the structural deformation and the environmental quality parameters in the multidimensional room inspection feature set are respectively compared with national standard values to obtain a parameter overrun grade index; The defect density, deformation deviation and environment parameter superscalar of the same space region are calculated, and a region quality deviation characteristic spectrum is constructed; Carrying out dimensionless treatment on the regional quality deviation characteristic spectrum to obtain a relative quality risk coefficient; Constructing a space quality coupling model based on the parameter overrun grade index and the relative quality risk coefficient; Performing multivariate correlation analysis on the surface defect characteristics, the structural deformation characteristics and the environmental impact characteristics to obtain a cross-modal correlation matrix; identifying key quality influence factors based on the cross-modal correlation matrix, and constructing a characteristic correlation network; and extracting risk characteristics according to the characteristic association network and the space quality coupling model to obtain a defect risk mapping network.
  5. 5. The room source intelligent room inspection method based on computer vision and multi-sensor fusion according to claim 4, wherein the dimensionless processing is performed on the regional quality deviation feature spectrum to obtain a relative quality risk coefficient, and the method comprises the following steps: Setting a quality deviation threshold interval, and carrying out normalization processing on the regional quality deviation characteristic spectrum to obtain a normalized deviation value; determining weight coefficients of surface defects, structural deformation and environmental parameters by adopting an entropy weight method; Constructing a weighted deviation matrix based on the standardized deviation value and the weight coefficient; And calculating the relative deviation ratio of each element in the matrix and the national standard value to obtain the relative quality risk coefficient.
  6. 6. The room source intelligent room inspection method based on the integration of computer vision and multiple sensors according to claim 1, wherein the building of the dynamic defect prediction model, and the evolution trend prediction and the severity assessment of the defect risk mapping network based on the dynamic defect prediction model, so as to obtain a room source quality risk map, comprises the following steps: Constructing a space-time attention mechanism model and an LSTM time sequence prediction model; fusing the space-time attention mechanism model with the LSTM model to construct an initial defect prediction model; Training an initial defect prediction model by using historical room inspection data to obtain a dynamic defect prediction model; Performing time sequence feature extraction on the defect risk mapping network by adopting a sliding time window method, and inputting the extracted features into a dynamic defect prediction model to obtain a defect evolution trend prediction result; constructing a quality degradation trend curve based on the defect evolution trend prediction result and the structural deformation characteristics; And comparing and analyzing the quality degradation trend curve with a preset safety level threshold value to obtain a house source quality risk map.
  7. 7. The intelligent room inspection method based on the integration of computer vision and multiple sensors according to claim 1, wherein the building of a room quality evaluation index system, the comprehensive grade evaluation of the room quality risk map, the acquisition of room inspection report of the room source, comprises the following steps: establishing a multidimensional quality evaluation index system comprising a defect severity index, a region association index, a trend deterioration index and an environment coupling index; Performing index quantification on the house source quality risk map based on a multidimensional quality evaluation index system to obtain a house source quality index; carrying out standardization treatment on each house source quality index to obtain a normalized quality index, and carrying out weighted fusion on the normalized quality index to obtain a comprehensive quality score; dividing house source quality grades based on comprehensive quality scores, and carrying out sensitivity analysis on the normalized quality indexes to obtain risk contribution indexes; and generating a room source room examination report according to the room source quality grade and the risk contribution index.
  8. 8. The intelligent room verification method based on computer vision and multisensor fusion of claim 7, wherein the weighted fusion of normalized quality indexes to obtain a comprehensive quality score comprises: Constructing a quality evaluation matrix, and mapping each normalized quality index into a quality risk probability; Performing membership calculation on the quality risk probability by using a fuzzy comprehensive evaluation method to obtain a fuzzy quality vector; Determining each index weight based on the risk contribution index, and carrying out weighted summation on fuzzy quality vectors; and mapping the weighted result to a [0,100] scoring interval to obtain the comprehensive quality score.
  9. 9. The room source intelligent room inspection method based on the integration of computer vision and multiple sensors according to claim 1, wherein the generating the rectification decision proposal according to the room source room inspection report comprises the following steps: Identifying a high risk area and key influence parameters based on a house source house inspection report, matching a historical modification case library, and deducing an optimal modification scheme; carrying out rectification urgency assessment according to the quality grade and the defect evolution trend prediction result to generate rectification priority; setting a correction window period suggestion based on the correction priority and the house source use plan; Generating a targeted correction process according to the defect type and the risk contribution index; And forming a rectification decision proposal based on the rectification window period proposal and the rectification process.
  10. 10. A room source intelligent room inspection system based on computer vision and multisensor fusion for implementing the room source intelligent room inspection method based on computer vision and multisensor fusion of any one of claims 1 to 9, the system comprising: The multi-mode sensing module is used for acquiring multi-mode sensing data of a room source space, and carrying out space-time registration and feature fusion on the multi-mode sensing data to acquire a multi-dimensional room inspection feature set; The defect analysis module is used for carrying out defect identification and parameter deviation analysis on the multidimensional room inspection feature set, constructing a space quality coupling model, and analyzing cross-modal feature relevance to obtain a defect risk mapping network; The risk prediction module is used for constructing a dynamic defect prediction model, and carrying out evolution trend prediction and severity assessment on the defect risk mapping network based on the dynamic defect prediction model so as to obtain a house source quality risk map; the quality evaluation module is used for establishing a house quality evaluation index system, and carrying out comprehensive grade evaluation on a house quality risk map to obtain a house verification report; And the decision pushing module is used for generating a correction decision suggestion according to the house source room inspection report and realizing real-time defect early warning and correction scheme pushing through edge calculation.

Description

House source intelligent house checking method and system based on computer vision and multi-sensor fusion Technical Field The invention relates to the field of intelligent house inspection, in particular to a house source intelligent house inspection method and system based on computer vision and multi-sensor fusion. Background With the acceleration of the urban process and the sustainable development of the real estate market, the house checking link in the house renting, trading and delivering processes is increasingly paid great attention to owners, tenants and property management parties. The traditional house inspection mainly relies on manual visual inspection and simple tool measurement, has the problems of strong subjectivity, low efficiency, non-uniform standard, easy missed inspection and the like, and is difficult to comprehensively and objectively evaluate the actual quality condition of the house. Especially in the scenes of precision house delivery, second-hand house transaction, long-leased apartment batch management and the like, the traditional mode has difficulty in meeting the requirements of efficient, accurate and intelligent house inspection. In recent years, the rapid development of computer vision, internet of things sensing and edge computing technologies provides a new technical path for intelligent laboratory tests. Part of researches try to acquire images by using cameras to identify cracks or monitor environmental states by using temperature and humidity sensors, but the problems of single data mode, insufficient information fusion, lack of space-time correlation analysis and the like generally exist, so that the defect identification precision is low, the risk assessment is on the one hand, and the dynamic and systematic assessment of the overall quality of a house is difficult to realize. In addition, the existing method focuses on static detection, lacks prediction capability for defect evolution trend, and cannot early warn potential structural potential safety hazards in advance. The foregoing is provided merely for the purpose of facilitating understanding of the technical solutions of the present invention and is not intended to represent an admission that the foregoing is prior art. Disclosure of Invention The invention mainly aims to provide a house source intelligent house checking method and system based on computer vision and multi-sensor fusion, and aims to solve the technical problems that the existing house checking technology depends on manpower, has single data mode, is insufficient in information fusion and lacks dynamic prediction capability, and is difficult to realize comprehensive, accurate and intelligent evaluation of house quality. The technical scheme for solving the technical problems is as follows: The invention provides a room source intelligent room inspection method based on computer vision and multi-sensor fusion, which comprises the steps of obtaining multi-mode sensing data of room source space, carrying out space-time registration and feature fusion on the multi-mode sensing data to obtain a multi-dimensional room inspection feature set, carrying out defect identification and parameter deviation analysis on the multi-dimensional room inspection feature set, constructing a space quality coupling model, analyzing cross-modal feature relevance to obtain a defect risk mapping network, constructing a dynamic defect prediction model, carrying out evolution trend prediction and severity assessment on the defect risk mapping network based on the dynamic defect prediction model to obtain a room source quality risk map, establishing a room source quality assessment index system, carrying out comprehensive grade assessment on the room source quality risk map to obtain a room source room inspection report, generating a rectification decision suggestion according to the room source room inspection report, and carrying out real-time defect early warning and rectification scheme pushing through edge calculation. The invention provides a room source intelligent room inspection system based on computer vision and multi-sensor fusion, which comprises a multi-mode sensing module, a defect analysis module, a risk prediction module, a quality evaluation module and a decision pushing module, wherein the multi-mode sensing module is used for acquiring multi-mode sensing data of a room source space, carrying out space-time registration and feature fusion on the multi-mode sensing data to acquire a multi-dimensional room inspection feature set, the defect analysis module is used for carrying out defect identification and parameter deviation analysis on the multi-dimensional room inspection feature set, constructing a space quality coupling model, analyzing cross-mode feature relevance to acquire a defect risk mapping network, the risk prediction module is used for constructing a dynamic defect prediction model, carrying out evolution trend prediction and severity eval