Sky Area Segmentation Using Multi-Scale CNN Feature Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for segmenting the sky area in images, such as those using gradient statistical techniques, have low accuracy in scenes with unclear sky boundaries like dense fog and dark nights, making them unsuitable for these conditions.
Innovation Solution
A convolutional neural network (CNN) is employed, comprising an image input layer, multiple cascaded convolutional neural networks, an up-sampling layer, and a sky area determining layer, which extracts and processes sky feature images of different scales to accurately identify the sky area by performing feature fusion and low-frequency feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If gradient statistical techniques are used for sky area segmentation, then the method is simple to implement, but the segmentation accuracy is low in scenes with unclear sky boundaries
Solution Approach 1:
The patent divides the sky area segmentation task into multiple processing stages: feature extraction, multi-scale feature fusion, and boundary refinement. The CNN architecture is segmented into multiple layers including convolutional layers, pooling layers, and fully connected layers, each performing specific functions to progressively improve segmentation accuracy while managing complexity through modular design
Solution Approach 2:
The patent introduces multi-scale feature extraction by processing images at different scales and dimensions. The CNN architecture operates across multiple spatial resolutions, fusing features from different scales to capture both local boundary details and global sky area characteristics, thereby improving accuracy without excessive complexity increase
2Measurement precision
If multi-scale feature extraction and fusion is performed, then the segmentation accuracy is improved, but the computational complexity increases
Solution Approach 1:
The computational task is divided into discrete CNN layers with specific functions: early layers extract basic features, middle layers perform multi-scale fusion, and later layers refine boundaries. This segmentation of computational work allows efficient resource utilization while achieving high accuracy through progressive feature processing
Solution Approach 2:
The CNN architecture performs multiple functions within a unified framework: feature extraction, multi-scale fusion, and boundary refinement are all achieved through the same network structure. This multi-functionality reduces overall computational overhead compared to using separate specialized systems for each task
Data Source
AI summary
The present disclosure provides a method and apparatus for segmenting a sky area, and a convolutional neural network. The method includes: acquiring, by the image input layer, an original image; extracting, by the first convolutional neural network, a plurality of sky feature images with different scales from the original image; processing, by the plurality of cascaded second convolutional neural networks, the plurality of sky feature images to output a target feature image; up-sampling, by the up-sampling layer, the target feature image to obtain an up-sampled feature image; determining, by the sky area determining layer, a pixel area of which a gray value is greater than or equal to a preset gray value in the up-sampled feature image as a sky area.