3D Virtual Model Pre-Processing for Object Removal and Hole Filling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating 3D virtual models often incorporate unwanted objects, such as people, into the model due to data collection issues, leading to noise and inaccuracies.
Innovation Solution
A pre-processing method using a deep learning-based artificial intelligence model to identify and delete unwanted objects from image data, followed by hole-filling using a neural network to maintain data integrity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data collection is performed by photographing several points in real space to create a 3D virtual model, then the virtual space implementation is achieved, but unwanted objects such as people are reflected in the model causing noise and inaccuracies
Solution Approach 1:
The patent extracts and removes unwanted objects (people, moving objects) from the collected image data before creating the 3D virtual model. This is achieved through object detection algorithms that identify and separate unwanted objects from the background environment, allowing the system to exclude these harmful elements from the final model construction process.
Solution Approach 2:
The patent applies preliminary processing to detect and remove unwanted objects before the 3D model creation process. By performing object detection and removal in advance (pre-processing stage), the system prevents noise from contaminating the model data from the beginning, rather than attempting to clean the data after model creation.
2Object-generated harmful factors
If unwanted objects are removed from image data by deleting data in corresponding areas, then noise is eliminated, but hole areas are created that distort the model
Solution Approach 1:
The patent uses copying techniques to fill hole areas by replicating information from surrounding regions. After unwanted objects are removed, the system copies texture, color, and structural information from adjacent valid pixels to reconstruct the missing areas, maintaining visual continuity and preventing model distortion.
Solution Approach 2:
The patent introduces an intermediary processing step (hole filling algorithm) between object removal and model creation. This intermediary process acts as a mediator that reconstructs missing data using surrounding context information, ensuring that the transition from object removal to complete image data is smooth and accurate.
3Device complexity
If traditional object removal methods are used without hole filling, then processing is simpler, but the 3D virtual model contains gaps and reduced quality
Solution Approach 1:
The patent implements self-service hole filling where the algorithm automatically analyzes surrounding image data and reconstructs missing areas without manual intervention. The system serves itself by detecting edge patterns, texture repetitions, and structural continuities to autonomously fill holes, maintaining high model quality while keeping the process automated and efficient.
Data Source
AI summary
A hole-filling method, performable on a computing device, for providing a three-dimensional virtual model according to a technical aspect of the present application may comprises the operations of: acquiring an original training image and a hole creation training image, wherein the hole creation training image is an image in which at least one hole is created based on the original training image; creating a hole-filling training image by performing hole-filling on the hole creation training image using a neural network; performing spherical transformation on each of the hole-filling training image and the original training images; and training the neural network based on the difference between the spherically transformed hole-filling training image and the spherically transformed original training image.