Application Scene Recognition Using Background Structure Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing scene recognition methods for application programs are limited in their ability to accurately recognize complex scenes due to limitations in training samples, models, and training algorithms, leading to inefficiencies in scene recognition capacity.
Innovation Solution
A method that extracts a first interface image from a recorded video of an application program, determines the background structure feature, and matches it with a preset scene's background structure feature to recognize the scene, using a simple image analysis algorithm without complex feature extraction, applicable to different platforms and models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional AI models with complex training are used for scene recognition, then recognition accuracy may be improved, but recognition speed and computational resources deteriorate
Solution Approach 1:
The patent segments the interface image into multiple regions (foreground region containing interface elements and background region) and extracts features separately from each region. This segmentation allows the system to focus on key discriminative features without processing the entire image, thereby improving recognition speed while maintaining accuracy.
Solution Approach 2:
The patent extracts specific background structure features from the interface image, such as background element types, their positions, and spatial relationships. By extracting only the necessary features rather than using complex AI models, the system achieves fast recognition without sacrificing accuracy.
2Reliability
If complex feature extraction methods are used, then scene recognition capability may be improved, but computational complexity and time consumption increase
Solution Approach 1:
The patent applies different feature extraction strategies to different regions of the interface image. The foreground region is analyzed for interface elements while the background region is analyzed for structural features. This localized approach simplifies the overall feature extraction process while maintaining comprehensive scene recognition capability.
Solution Approach 2:
The patent changes the parameters being extracted from complex deep learning features to simpler structural parameters such as background element types, positions, and spatial relationships. This parameter transformation reduces computational complexity while preserving the essential information needed for reliable scene recognition.
3Measurement precision
If comprehensive interface elements are analyzed for scene recognition, then recognition accuracy may be improved, but processing time increases
Solution Approach 1:
The patent performs preliminary analysis by first identifying and segmenting the background region, then extracts background structure features in advance. This preliminary action allows the system to have feature information ready for quick comparison and matching, reducing the overall processing time while maintaining accurate scene recognition.
Data Source
AI summary
Some aspects of the disclosure provide a method of scene recognition. In some examples, a first video that includes interface images of a computer device recorded when an application program is executed on the computer device can be obtained. A first interface image is extracted from the first video, the first interface image includes a first interface element area with one or more preset interface elements of the application program, and a first background area excluding the first interface element area. A first background structure feature of the first background area in the first interface image is determined. Whether the first background structure feature satisfies a matching condition to a second background structure feature of a preset scene is determined. The first interface image is determined to correspond to the preset scene when the first background structure feature satisfies the matching condition to the second background structure feature.