Image Processing Likelihood Calculation for Target Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately detecting target objects due to the need for manual adjustments of detection parameters, leading to false positives and false negatives, and rely solely on image features specific to the detection algorithm, which can result in incorrect object detection.
Innovation Solution
An image processing apparatus and method that utilizes a machine learning device to perform learning on partial images from input images, calculating a likelihood of target object detection based on detection results, allowing for automatic determination of detection parameters and incorporating features beyond those used in the detection algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If the threshold value of the degree of correspondence is set low to detect more target objects, then the detection coverage is improved, but false positive detection increases
Solution Approach 1:
The patent introduces a likelihood calculation mechanism as an intermediary between the detection algorithm and the final detection result. The machine learning device calculates a likelihood value that represents the probability of correct detection, which serves as a mediator to evaluate detection results beyond the simple degree of correspondence threshold, thereby reducing false positives while maintaining detection coverage
Solution Approach 2:
The system implements feedback by using the calculated likelihood values to adjust and optimize detection parameters. The machine learning device learns from detection results and feedback information to improve the accuracy of likelihood calculation, which in turn refines the detection process and reduces false positives over time
2Reliability
If the threshold value of the degree of correspondence is set high to reduce false positives, then detection reliability is improved, but false negative detection increases
Solution Approach 1:
The patent changes the parameter used for detection decision from solely relying on the degree of correspondence to incorporating the likelihood value calculated by the machine learning device. This parameter change allows the system to maintain high detection accuracy while improving detection coverage, as the likelihood calculation captures additional information about detection confidence that the degree of correspondence alone cannot provide
3Measurement precision
If manual adjustment of detection parameters is performed through trial and error to optimize detection, then detection precision can be improved, but the complexity of operation increases
Solution Approach 1:
The system implements self-service by automatically optimizing detection parameters through the machine learning device. The device learns from training data and automatically determines optimal detection parameters without requiring manual trial and error adjustment, thereby maintaining high detection precision while significantly reducing operational complexity
Solution Approach 2:
The system uses feedback from detection results to automatically adjust and optimize detection parameters. The machine learning device learns from the feedback of detection outcomes and iteratively improves parameter settings, eliminating the need for manual trial and error while maintaining optimal detection precision
4Device complexity
If only image features used in the detection algorithm are considered for detection, then the detection process is simplified, but detection accuracy decreases due to algorithm-specific limitations
Solution Approach 1:
The machine learning device serves multiple functions: it not only calculates likelihood values but also learns to recognize patterns that improve detection accuracy. By incorporating additional image features and contextual information beyond what the detection algorithm uses, the system achieves more accurate detection without significantly increasing overall process complexity
Solution Approach 2:
The likelihood calculation acts as an intermediary layer that processes and integrates multiple features including those beyond the detection algorithm's scope. This intermediary mechanism synthesizes information from various sources to provide more accurate detection results while maintaining a relatively simple overall system architecture
Data Source
AI summary
An image processing apparatus, which receives an input image and detects an image of a target object based on a detection algorithm, includes a machine learning device which performs learning by using a plurality of partial images cut out from at least one input image, based on a result of detection of the image of the target object, and calculates a likelihood of the image of the target object.