Image Classification via Generator Features and Cost Functions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Advanced driver assistance systems face high false positive detections and misdetection probabilities when classifying objects in images, which hampers effective identification of road users and accident avoidance.
Innovation Solution
The system employs a method of image classification based on generator features, where elements are clustered using a centroid-based algorithm, and a cost function is computed for each element to determine its belonging to a cluster, reducing false positives and misclassifications by adjusting the number of clusters and recalculating features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If classification is based directly on elements of the cluster (distance between elements), then the classification process is simple, but the false positive detections and misdetection probability increase
Solution Approach 1:
The patent transforms the classification approach by changing the parameters used for classification. Instead of using raw element distances directly, it computes cost functions that measure the deviation of each element from the cluster centroid in a transformed feature space. This parameter transformation enables more accurate separation of true objects from false positives while maintaining computational feasibility through established algorithms like k-means and SVM.
2Reliability
If the number of clusters is increased to reduce misclassification, then the detection accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent applies segmentation by dividing the classification problem into distinct stages: (1) initial clustering of elements into K groups using k-means, (2) computation of cost functions for each element relative to its cluster centroid, (3) application of SVM classification on the cost function values, and (4) final object detection based on classified clusters. This segmented approach allows systematic control of computational complexity while achieving high classification accuracy.
Solution Approach 2:
The patent performs preliminary clustering using k-means algorithm before applying the final SVM classification. This preliminary action groups similar elements together, reducing the search space and enabling more efficient final classification. The cost function computation based on cluster centroids serves as a preprocessing step that simplifies the subsequent binary classification task.
Data Source
AI summary
A method including receiving, by an image classification engine, a number K of clusters to be created by a clustering algorithm. The method further including receiving, by the image classification engine, a set of elements based on an image, executing, by the image classification engine, the clustering algorithm on the set of elements to create K clusters, each cluster having a respective subset of the set of elements, for each cluster, computing, by the image classification engine, a centroid of the cluster, for each cluster, creating, by the image classification engine, a generator of the cluster based on the respective subset of the set of elements corresponding to each cluster, and for each element of each cluster, computing, by the image classification engine, a cost function corresponding to the element based on the centroid, the respective subset of the set of elements, and the generator corresponding to the cluster.