Neural Network Occupant Classification Using Static-Dynamic State Machine
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vision-based occupant classification systems face challenges in achieving accurate and rapid dynamic classification, particularly in distinguishing between adult and child positions, due to complexities in image features and algorithms, leading to false detections and inadequate response times during vehicle accidents.
Innovation Solution
A neural network engine with a state machine is employed to classify occupant changes between adults and children, utilizing wavelet features and novel differential edge density features from stereo images, switching between dynamic and static classification processes to enhance accuracy and speed, and prevent false detections.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex image features and algorithms are used for vision-based dynamic classification, then classification accuracy is improved, but processing speed deteriorates and fails to meet the required response time of a few tens of milliseconds
Solution Approach 1:
The patent divides the classification task into two distinct phases: static classification (performed before occupancy detection) and dynamic classification (performed during occupancy detection). The static classifier establishes prior knowledge about occupant type using complex algorithms, while the dynamic classifier uses this prior knowledge to perform faster real-time classification. This segmentation allows complex processing to be done in advance, meeting the millisecond response time requirement during actual occupancy events.
Solution Approach 2:
The patent performs static classification in advance to establish prior knowledge of occupant type before dynamic classification is needed. This preliminary action includes determining whether the seat is occupied and classifying the occupant type using comprehensive image features and algorithms. When dynamic classification becomes necessary, the system already has this baseline information, allowing for faster processing and meeting the required response time of a few tens of milliseconds.
2Speed
If motion-detection based state machine is used for dynamic classification, then processing speed is improved, but false detections occur under certain conditions such as slow occupant entry/exit or specific movement patterns
Solution Approach 1:
The patent incorporates feedback mechanisms where the dynamic classifier's output is validated against the static classifier's prior knowledge. When the dynamic classifier produces a classification, it is cross-checked with the established occupant type from static classification. This feedback loop helps eliminate false detections by comparing real-time motion-based classification with the baseline knowledge, ensuring higher reliability while maintaining fast processing speed.
Solution Approach 2:
The patent merges the strengths of both static and dynamic classification approaches into a unified system. The static classifier provides accurate prior knowledge establishment, while the dynamic classifier provides fast real-time response. By combining these two classifiers and using the dynamic classifier's output to update or validate the static classification results, the system achieves both high speed and high reliability, eliminating the false detection problems of motion-detection alone.
3Speed
If a cascaded classification structure with prior knowledge from static classification is used, then processing speed is improved, but the system complexity increases
Solution Approach 1:
The patent segments the classification system into distinct functional modules: static classification module, dynamic classification module, and a coordination layer that manages the cascaded structure. Each module has a specific responsibility - static classification establishes prior knowledge, dynamic classification performs fast real-time classification, and the coordination layer manages data flow between them. This modular segmentation makes the complex system more manageable and implementable while maintaining the speed benefits of the cascaded structure.
Data Source
AI summary
A method of classifying vehicle occupants utilizes a neural network engine having a state machine for determining if an occupant has changed preferably between an adult and adult, an adult and child, and a child and child from a pair of images. If no change has occurred, the method utilizes the prior occupant type and then decides if the occupant has changed in position. If no, the occupant is deemed static and the prior type is valid as a classification or output to preferably a vehicle restraint system. If the occupant has changed in position, a dynamic classification process is initiated by either an adult or a child dynamic classifier as dictated by the state machine. Valid dynamic classifier outputs or classifications can be sent to the restraint system and invalid dynamic classifier outputs are sent to a static classifier for update of the occupant type.