Neural Network Size Identification from Fused Object Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods require imaging a reference object with the measurement target object or acquiring camera characteristics to estimate size, which is inefficient and limits accuracy.
Innovation Solution
A method and device using an artificial neural network to identify the size of a measurement target object by imaging a reference object, acquiring its image, fusing it with the target object image, and inputting the fused images to a neural network model to determine the size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a distance sensor or reference object is used to estimate size, then size estimation can be performed, but the process becomes complex and requires additional equipment
Solution Approach 1:
The patent extracts and removes the reference object from the imaging process, using only the target object's image characteristics (perspective distortion, relative positions of features) to determine size. This eliminates the need for additional reference objects or distance sensors while maintaining size estimation capability through neural network analysis of the single target object image
Solution Approach 2:
The target object itself provides all necessary information for size determination without requiring external reference objects or additional sensors. The neural network extracts size information directly from the target object's image features including perspective distortion and relative feature positions, making the system self-sufficient
2Measurement precision
If camera characteristic information is acquired in advance, then size identification accuracy improves, but the preparation time and complexity increase
Solution Approach 1:
The neural network is pre-trained with general knowledge about object sizes and perspectives, but no specific camera characteristics need to be acquired beforehand. The system performs preliminary learning during training phase, then can immediately process target objects without additional calibration or camera parameter acquisition, eliminating preparation time while maintaining accuracy
Solution Approach 2:
The system changes from requiring fixed camera parameter inputs to using variable image features (perspective distortion, relative positions) that adapt to different camera configurations automatically. The neural network learns to interpret these varying parameters dynamically without needing predetermined camera characteristics
3Measurement precision
If reference object imaging is performed together with target object, then size measurement becomes possible, but the imaging process and data processing become more complex
Solution Approach 1:
The reference object is completely removed from the imaging process. The system uses only the target object's image, extracting size information from its perspective distortion and internal feature relationships. This single-object imaging approach eliminates the complexity of coordinating multiple objects while maintaining measurement capability
Solution Approach 2:
The neural network serves multiple functions using only the target object image: it simultaneously determines size, accounts for perspective distortion, and identifies relative positions without needing separate reference objects. This multi-functionality is achieved through comprehensive training on diverse image characteristics
Data Source
AI summary
Provided are a method and electronic device for identifying a size of a measurement target object. The method includes imaging a reference object, which is a reference for identifying the size of the measurement target object, to acquire a reference object image, imaging the measurement target object to acquire a target object image, fusing the acquired reference object image and the acquired target object image, and inputting the fused reference object image and target object image to a first neural network model to acquire size information of the measurement target object from the first neural network model.