Environmental Image Perception Using Segmented Multi-Model Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis and processing technologies face challenges in achieving accurate and efficient perception of multiple target classes in environmental images, which affects the overall accuracy of image analysis and processing, particularly in applications like self-driving.
Innovation Solution
The method involves segmenting a raw environmental image based on different perception target classes and using corresponding perception models to adjust and perceive the image, followed by fusing the perception results to improve accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single unified perception model is used to process the entire environmental image, then the device complexity is reduced, but the measurement precision and perception accuracy deteriorate
Solution Approach 1:
The environmental image is segmented into multiple regions corresponding to different perception target classes (vehicles, pedestrians, cyclists, animals, obstacles). Each region is processed by a dedicated perception model optimized for that specific target type, thereby improving perception accuracy without requiring a single overly complex unified model.
Solution Approach 2:
Different perception models with specialized characteristics are applied to different regions of the image based on the local requirements of each perception target class. Each model is tailored to detect specific features relevant to its target class, achieving high precision through localized optimization rather than a general-purpose approach.
2Measurement precision
If multiple specialized perception models are used for different target classes, then the measurement precision improves, but the device complexity increases
Solution Approach 1:
The image processing task is divided into multiple independent sub-tasks, each handled by a specialized model. This segmentation allows each model to be simpler and more focused, reducing the complexity burden on individual models while collectively achieving high overall accuracy through the division of labor.
Solution Approach 2:
The system employs multiple specialized models that collectively cover all perception target classes. Each model is designed for a specific function (detecting a particular target type), and together they provide universal coverage for the entire environmental scene, achieving both specialization and comprehensiveness.
3Productivity
If the environmental image is processed without segmentation, then the processing speed is higher, but the measurement precision deteriorates
Solution Approach 1:
The image is segmented into multiple regions of interest based on perception target classes. By processing only relevant regions with specialized models rather than the entire image uniformly, the system maintains high processing speed while improving accuracy through focused analysis of critical areas.
Solution Approach 2:
Instead of applying full processing to the entire image, the system applies specialized perception actions only to segmented regions containing specific target types. This partial action approach concentrates computational resources where needed, maintaining speed while enhancing precision for each target class.
4Measurement precision
If image segmentation and multiple model processing are applied, then the measurement precision improves, but the loss of time increases
Solution Approach 1:
The image processing is segmented into parallel independent tasks, each handled by a specialized model. This allows simultaneous processing of different target classes, reducing overall processing time while maintaining high accuracy through dedicated models for each segment.
Solution Approach 2:
The image is pre-segmented into regions corresponding to different perception target classes before model processing. This preliminary organization enables more efficient processing by the specialized models, reducing the time required for accurate detection compared to processing the entire image without pre-segmentation.
Data Source
AI summary
The present disclosure provides an image processing method and apparatus, a system, a medium, a device and a driving system, which belong to the field of data processing technologies. The image processing method mainly includes: according to perception target features of each perception target feature class in a plurality of predetermined perception target feature classes, and a pixel requirement of a neural network image perception model corresponding to each perception target feature class in a plurality of predetermined neural network image perception models, performing corresponding adjustment to a current environmental image to obtain a plurality of current environmental feature images; by using the corresponding neural network image perception model, perceiving a perception target in each current environmental feature image to obtain a plurality of perception results; and fusing the plurality of perception results to obtain perception result data of the current environmental image. The present disclosure can improve perception accuracy.