Multi-class Object Classification via Linear Mapping and Overlap Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-class object classification methods face inefficiencies due to complex calculations and low accuracy in binary classification, especially when dealing with numerous classes and near-distance objects, leading to prolonged processing times and reduced classification efficiency.
Innovation Solution
A multi-class object classification method that involves receiving and processing training images to generate classifiers using a linear mapping method, dividing characteristic images into blocks, and selecting variances to determine parameter ranges and thresholds, allowing for efficient classification by recording overlapping ranges and selecting appropriate classifiers to reduce classification steps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multi-class classification is performed directly using boosting algorithm, then all representative points of input images are analyzed to classify target objects, but the calculation becomes more complicated and time-consuming with lower accuracy
Solution Approach 1:
The patent segments the multi-class classification problem into multiple binary classification problems. Instead of analyzing all representative points simultaneously through a complex multi-class classifier, the system divides the task into multiple simpler binary classifiers, each handling a specific class distinction. This segmentation reduces the computational complexity and processing time while maintaining classification accuracy.
2Device complexity
If binary classification is used to solve multi-class problems, then the classification process can be simplified, but there are too many pending characteristic points of images and the calculation time becomes excessively long when different objects are in near distance
Solution Approach 1:
The patent extracts and removes redundant or less important characteristic points from the image data before classification. By taking out only the most relevant feature points and eliminating unnecessary calculations, the system reduces the number of pending characteristic points that need to be processed, thereby decreasing calculation time while maintaining the simplified binary classification approach.
3Adaptability or versatility
If the number of classes is huge, then the number of classification processes has to increase to distinguish every class, but this results in lower efficiency of the classification
Solution Approach 1:
The patent merges multiple binary classification results to achieve multi-class classification. Instead of creating a separate classification process for each class, the system combines the outcomes of multiple binary classifiers working in parallel or sequence. This merging approach allows the system to handle a huge number of classes efficiently, as the binary classifiers can be processed simultaneously and their results integrated, thereby maintaining high classification efficiency despite the large number of classes.
Data Source
AI summary
A multi-class object classifying method and system are disclosed herein, where the multi-class object classifying method includes the following steps: classes, first training images and second training images are received and stored, and first characteristic images and second characteristic images are respectively extracted from the first training images and the second training images; the first training images is used to generate classifiers through a linear mapping classifying method; a classifier and the second characteristic images are used to determine parameter ranges corresponding to the classes and a threshold corresponding to the classifier. When two of the parameter ranges overlap, the remaining parameter ranges except for the two overlapped parameter ranges are recorded; after another classifier is selected from the classifiers except for the classifier that has been selected, the previous steps is repeated until the parameter ranges don't overlap with each other and the parameter ranges are recorded.