Vehicle Occlusion Detection via Dual Classifier Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection frameworks for autonomous vehicles fail to accurately and efficiently detect occlusions, which is crucial for safe motion planning and operation, as they cannot recover the shape of occluded objects or effectively handle occlusion detection in complex traffic environments.
Innovation Solution
A system and method for vehicle occlusion detection that involves training two machine learning classifiers: one for static images and another for image sequences, using extracted features to determine the occlusion status of vehicles, allowing for accurate tracking and classification of occlusion types such as occluding, being occluded, or being between other objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current object detection frameworks are used, then basic object detection is achieved, but occlusion detection accuracy is insufficient
Solution Approach 1:
The patent segments the detection task into two distinct classifiers: a first classifier for objects that cannot be tracked back to previous frames (including occluded objects), and a second classifier for objects that can be tracked. This segmentation allows each classifier to be specialized for its specific detection scenario, improving overall occlusion detection accuracy while maintaining system reliability.
2Measurement precision
If a single classifier is used for all objects, then device complexity is reduced, but detection accuracy for occluded objects deteriorates
Solution Approach 1:
The patent applies local quality by using different classifiers for different object types based on their tracking history. Objects that cannot be tracked back to previous frames are processed by the first classifier, while tracked objects use the second classifier. This localized specialization improves detection accuracy for each category without requiring a single overly complex universal classifier.
3Stability of the object's composition
If object tracking is performed for all frames, then temporal consistency is improved, but processing time increases
Solution Approach 1:
The patent implements partial action by applying the computationally intensive second classifier only to objects that can be tracked back to previous frames, while using the faster first classifier for objects that cannot be tracked. This selective approach maintains temporal consistency for trackable objects while reducing overall processing time by avoiding redundant classification of untrackable objects.
Data Source
AI summary
A system and method for vehicle occlusion detection is disclosed. A particular embodiment includes: receiving training image data from a training image data collection system; obtaining ground truth data corresponding to the training image data; performing a training phase to train a plurality of classifiers, a first classifier being trained for processing static images of the training image data, a second classifier being trained for processing image sequences of the training image data; receiving image data from an image data collection system associated with an autonomous vehicle; and performing an operational phase including performing feature extraction on the image data, determining a presence of an extracted feature instance in multiple image frames of the image data by tracing the extracted feature instance back to a previous plurality of N frames relative to a current frame, applying the first trained classifier to the extracted feature instance if the extracted feature instance cannot be determined to be present in multiple image frames of the image data, and applying the second trained classifier to the extracted feature instance if the extracted feature instance can be determined to be present in multiple image frames of the image data.