CN-121170153-B - Dynamic updating system for live-action three-dimensional model

CN121170153BCN 121170153 BCN121170153 BCN 121170153BCN-121170153-B

Abstract

The invention discloses a live-action three-dimensional model dynamic updating system, which particularly relates to the technical field of three-dimensional model dynamic updating, and comprises the steps of acquiring a candidate change region through a depth change detection network based on a graph attention mechanism, combining two high-precision evaluation indexes of a redundancy reconstruction abnormal coefficient and a structure damage pollution coefficient through incremental partial reconstruction and preliminary insertion simulation, quantifying the reconstruction effect and the insertion quality in real time, accurately capturing potential structure continuity damage risks, constructing a model update degradation objective function and degradation index evaluation mechanism, enabling the system to quantitatively monitor the risk of each round of update behaviors, timely identifying the closed loop trend of 'misjudgment-redundancy update-structure pollution-re misjudgment', and finally realizing automatic freezing and to-be-inspected marking of the high-risk candidate region and continuous pushing of a safety update path based on a fluctuation control strategy of the update degradation index, and guaranteeing seamless connection of model geometry and texture.

Inventors

WANG JINLIANG
GUAN JINSHENG
LI YANG

Assignees

湖南创信伟立科技股份有限公司

Dates

Publication Date: 20260512
Application Date: 20250923

Claims (3)

1. The dynamic updating system for the live-action three-dimensional model is characterized by comprising a region sensing module, a region reconstruction insertion module, a reconstruction destruction module, an updating degradation evaluation module and an updating decision control module; The area sensing module is used for acquiring a multi-source live-action data set through multiple channels, preprocessing the multi-source live-action data set and acquiring a candidate change area by adopting a depth change detection network based on a graph attention mechanism; The regional reconstruction insertion module is used for executing incremental local three-dimensional reconstruction and local reconstruction model insertion simulation on the candidate change region; The reconstruction damage module is used for acquiring redundant reconstruction information of the candidate change region and structural consistency information inserted by the local reconstruction model, wherein the redundant reconstruction information comprises a redundant reconstruction abnormal coefficient, and the structural consistency information comprises a structural damage pollution coefficient; The update degradation evaluation module is used for constructing an objective function of model update degradation according to the redundancy reconstruction abnormal coefficient and the structure damage pollution coefficient, obtaining a model update degradation index and evaluating degradation risk in a model update process; the updating decision control module is used for controlling and optimizing the updating decision according to the degradation risk of the model updating process; Constructing a model update degradation objective function according to the redundancy reconstruction abnormal coefficient and the structural damage pollution coefficient, obtaining a model update degradation index, and obtaining the model update degradation objective function as follows In the following The degradation index is updated for the model and, For the redundant reconstruction of the outlier coefficients, In order to destroy the contamination coefficient of the structure, Respectively representing the preset proportion coefficients of the redundancy reconstruction abnormal coefficient and the structural damage pollution coefficient, and Are all greater than 0; and controlling and optimizing the updating decision according to the degradation risk of the model updating process, wherein the method comprises the following steps: Obtaining a model update degradation index after each round of model update according to an objective function of model update degradation to construct an update degradation risk set: wherein The degradation index is updated for the model after the model update of the d-th round, Updating the total round for the model; updating fluctuation value of degradation index based on update degradation risk set calculation model : Wherein The average of the degradation index is updated for the model, ; Comparing the fluctuation value of the model update degradation index with a preset fluctuation threshold value, and performing control optimization on the update decision, wherein the control optimization is as follows: If the fluctuation value of the model update degradation index is larger than the fluctuation threshold, marking a candidate change area which is currently participated in the local three-dimensional reconstruction and the local reconstruction model insertion simulation as an area to be checked, simultaneously freezing update operation, and recording abnormal metadata for manual review; If the fluctuation value of the model update degradation index is smaller than or equal to the fluctuation threshold value, the current update strategy is continuously executed.
2. The dynamic update system of a live-action three-dimensional model of claim 1, wherein the redundant reconstruction anomaly coefficient acquisition logic is as follows: Parallax tensor of candidate change regions is extracted by parallax matching of M view image pairs corresponding to each candidate change region ; Extracting the variation amplitude of parallax under different visual angles to obtain parallax fluctuation values ; Mapping parallax fluctuation values to spatial disturbance indexes ; Recording probability response values detected as candidate change areas at T historical moments, and constructing a probability response value sequence; calculating the degree of polarization of the regions detected as candidate changes ; Mapping polarization degree to time fluctuation response index ; Extracting a point cloud set from the reconstruction of the candidate change region; for each point cloud Is a neighborhood of (a) Calculating a local covariance matrix ; For local covariance matrix Solving eigenvalues, and sorting from big to small Calculating local degradation ratio : Wherein Respectively represent local covariance matrices Mapping local degradation ratio to regional geometric degradation index : ; Calculating redundant reconstruction anomaly coefficients according to the spatial disturbance index, the time fluctuation response index and the regional geometric degradation index : Wherein A preset proportionality coefficient respectively representing the spatial disturbance index, the time fluctuation response index and the regional geometrical degradation index, and Are all greater than 0.
3. The dynamic update system of a live-action three-dimensional model according to claim 1, wherein the acquisition logic of the structural damage pollution coefficient is as follows: reconstructing a local redundancy model And original global three-dimensional model Is marked as an insertion interface of (a) Obtaining the boundary point set , For the number of boundary points, each boundary point Carrying normal vector Curvature of Spatial position vector ; For each boundary point Calculating normal disturbance : Obtaining a discrete disturbance vector field; For each boundary point Calculating curvature disturbance : Obtaining a discrete curvature disturbance field; constructing a spatial disturbance propagation map based on a set of boundary points Wherein , Acquiring edge weights for edge sets based on spatial location vectors ; Taking the normal disturbance and curvature disturbance as input signals, and executing disturbance propagation convolution operation: ; Obtaining global maximum disturbance concentration : ; Calculating spatial discrete values of disturbance concentration : Wherein In order to average the concentration of the disturbance, ; Calculating the structural damage pollution coefficient : 。

Description

Dynamic updating system for live-action three-dimensional model Technical Field The invention relates to the technical field of dynamic updating of three-dimensional models, in particular to a dynamic updating system of a live-action three-dimensional model. Background With the wide application of the live-action three-dimensional modeling technology in the fields of digital twin cities, digital cultural relics protection, intelligent operation and maintenance and the like, higher requirements are put forward on the dynamic updating capability of the three-dimensional model. In order to improve timeliness and reality consistency of the model, the prior art mostly adopts an incremental updating method, namely, urban or scene data are periodically acquired through a multi-source sensor, and local reconstruction and insertion updating are carried out on a changed area. However, in the actual operation process, the system often faces the superposition risk of misjudgment of a change area and structural logic destruction, and the problem of coupling chain of structural pollution and false evolution which is ignored for a long time is formed. Specifically, when the current change detection module processes factors such as illumination change, dynamic shielding (such as pedestrians, vehicles, tree shadows) and data acquisition inconsistency, the current change detection module often misjudges the unstructured short-term change as a structural change, so as to trigger unnecessary update operation. In the process of executing redundancy reconstruction and inserting the local model into the original global three-dimensional model, if the structural consistency is not maintained, the geometric continuity of the original model is easily damaged, so that the problems of fault, floating, penetration and the like are caused. More seriously, the structural pollution is often identified as 'new change' by a change detection module in a subsequent period again, repeated updating is caused, and an update degradation closed loop of 'misjudgment-redundancy update-structural pollution-misjudgment again' is formed. Therefore, a method for dynamically updating a real three-dimensional model with variation perception robustness, structural consistency guarantee capability and pollution inhibition mechanism is needed to effectively break the degradation coupling chain and improve the overall stability of the system and the updating quality of the model. Disclosure of Invention In order to overcome the above-mentioned drawbacks of the prior art, an embodiment of the present invention provides a dynamic update system for a live-action three-dimensional model, so as to solve the problems set forth in the above-mentioned background art. In order to achieve the above purpose, the present invention provides the following technical solutions: a dynamic updating system of a live-action three-dimensional model comprises a region sensing module, a region reconstruction insertion module, a reconstruction destruction module, an updating degradation evaluation module and an updating decision control module; The area sensing module is used for acquiring a multi-source live-action data set through multiple channels, preprocessing the multi-source live-action data set and acquiring a candidate change area by adopting a depth change detection network based on a graph attention mechanism; The regional reconstruction insertion module is used for executing incremental local three-dimensional reconstruction and local reconstruction model insertion simulation on the candidate change region; The reconstruction damage module is used for acquiring redundant reconstruction information of the candidate change region and structural consistency information inserted by the local reconstruction model, wherein the redundant reconstruction information comprises a redundant reconstruction abnormal coefficient, and the structural consistency information comprises a structural damage pollution coefficient; The update degradation evaluation module is used for constructing an objective function of model update degradation according to the redundancy reconstruction abnormal coefficient and the structure damage pollution coefficient, obtaining a model update degradation index and evaluating degradation risk in a model update process; and the updating decision control module is used for controlling and optimizing the updating decision according to the degradation risk of the model updating process. In a preferred embodiment, the acquisition logic for the redundant reconstruction anomaly coefficients is as follows: Parallax tensor of candidate change regions is extracted by parallax matching of M view image pairs corresponding to each candidate change region ; Extracting the variation amplitude of parallax under different visual angles to obtain parallax fluctuation values; Mapping parallax fluctuation values to spatial disturbance indexes; Recording probability response va