Scene Recognition via Object Detection Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional scene recognition methods are limited to recognizing fixed classes of scenes and require manual updates for new scene requirements, leading to high costs, low efficiency, and low precision.
Innovation Solution
A scene recognition method that uses an object detection model comprising a backbone network, feature pyramid network, convolutional network, and prediction network to recognize scenes by inputting images and object images, allowing for quick and precise recognition of any scene without manual intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual scene recognition is performed for new scene requirements, then scene recognition precision can be maintained, but recognition efficiency decreases and costs increase
Solution Approach 1:
The object detection model performs automatic scene recognition without requiring manual intervention. The system extracts features from images, detects objects automatically, and determines scene types through algorithmic processing, enabling the system to serve itself rather than requiring human operators for each recognition task.
Solution Approach 2:
The patent replaces manual mechanical scene recognition with an automated computer vision system. The object detection model uses neural networks and algorithmic processing to substitute human operators, achieving both high precision and improved efficiency through automated image analysis and object detection.
2Device complexity
If fixed-class scene recognition is implemented, then system complexity remains low, but adaptability to new scenes deteriorates
Solution Approach 1:
The object detection model serves multiple functions: it detects various objects (cones, barrels, signboards, vehicles), identifies different scene types (construction, traffic, accident, normal), and adapts to new scene requirements. This universal model replaces the need for separate fixed-class recognition systems, achieving high adaptability without proportionally increasing complexity.
Solution Approach 2:
The system transitions from static fixed-class recognition to dynamic scene understanding. The object detection model can identify new scene types by detecting combinations of objects and their spatial relationships, allowing the system to adapt dynamically to new scenarios without requiring pre-programmed categories for every possible scene.
3Measurement precision
If manual scene recognition is performed, then recognition precision can be maintained, but time consumption increases
Solution Approach 1:
The object detection model processes images continuously and automatically, maintaining high recognition precision through consistent algorithmic application. The system performs real-time object detection and scene classification without interruption or manual review, eliminating the time loss associated with manual recognition while preserving accuracy through automated feature extraction and analysis.
Data Source
Figure 1~3
Figure 4~6
AI summary
The disclosure relates to the field of image recognition technologies, and specifically provides a scene recognition method, a method for obtaining scene data, a device, a medium, and a vehicle, to resolve technical problems of low scene recognition efficiency and low recognition precision of an existing scene recognition method. The scene recognition method of the disclosure includes: obtaining an image to be recognized and at least one object image; constructing and training an object detection model; and recognizing a current scene based on the object detection model, the image to be recognized, and the at least one object image, to obtain a scene recognition result. In this way, any scene can be quickly recognized, and a scene recognition result with high precision can be obtained.