Image Processing Device Heterogeneous Target Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing techniques face limitations in enhancing detection accuracy for targets in images, as increasing the number of detection processes to improve accuracy leads to increased processing time and resource usage, with a cap on the achievable accuracy.
Innovation Solution
An image processing device and method that includes a detection target detection unit, a heterogeneous target determination unit, and a detection target determination unit to differentiate between the desired detection target and potential heterogeneous targets, reducing false detections by determining the class and number of detection classes based on certainty degrees and weighting values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of types of detection processes is increased to enhance detection accuracy, then detection accuracy is improved, but detection process time and resource amount increase in proportion
Solution Approach 1:
The detection process is segmented into multiple specialized detection units (face detection unit, ear detection unit, eye detection unit, etc.), each responsible for detecting specific features. This allows the system to perform targeted detection rather than exhaustive multi-type detection, reducing overall processing time while maintaining accuracy.
Solution Approach 2:
The system performs preliminary detection with a first detection process to identify candidate targets, then applies a second detection process only to specific regions or candidates. This staged approach avoids running all detection processes on the entire image, significantly reducing detection time while preserving accuracy for critical targets.
2Measurement precision
If the number of types of detection processes is increased to enhance detection accuracy, then detection accuracy is improved, but resource amount of memory and processing power increases
Solution Approach 1:
Detection functionality is divided into separate modular units (face detection unit, ear detection unit, eye detection unit, etc.), allowing the system to load and execute only the necessary detection modules based on detection needs, optimizing memory usage and resource allocation.
Solution Approach 2:
The system applies detection processes selectively rather than universally - using a first detection process for initial identification and a second detection process only where needed. This partial application of detection resources reduces overall resource consumption while achieving sufficient accuracy for the detection task.
3Productivity
If conventional detection processes are used to detect targets, then detection speed is maintained, but false detections occur reducing accuracy
Solution Approach 1:
A region of interest determination unit acts as an intermediary between the first detection process and the second detection process. It analyzes the first detection results to identify and define regions of interest, then directs the second detection process to focus on these specific regions. This intermediary step filters out false detection areas and concentrates processing power on likely targets, improving accuracy without sacrificing overall detection speed.
Data Source
AI summary
An image processing device configured to detect a detection target, which is all or a part of a predetermined main body on an image has a detection target detection unit that detects an estimated detection target that the image processing device assumes to be the detection target from the image, a heterogeneous target determination unit that determines whether the estimated detection target detected by the detection target detection unit is an estimated heterogeneous target that the image processing device assumes to be a heterogeneous target, which is all or a part of a main body different in class from the main body, and a detection target determination unit that determines whether the estimated detection target detected by the detection target detection unit is the detection target based on a determination result of the heterogeneous target determination unit.