Pedestrian Detection via Optical Flow Confidence Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current pedestrian detection methods in motor vehicles require significant computational power and are costly, especially in embedded systems, and often rely on pattern recognition or stereo cameras, which are resource-intensive and less commonly used.
Innovation Solution
A method that uses optical flow vectors from a temporal sequence of images to determine confidence metrics, allowing for pedestrian detection without pattern recognition, using predefined plausibility criteria and threshold values, which can be efficiently processed in embedded systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pattern recognition based on characteristic features is used for pedestrian detection, then detection reliability is improved, but computational power requirement increases significantly
Solution Approach 1:
The patent extracts and uses only the necessary motion information from optical flow vectors (direction and magnitude) without performing full pattern recognition. It takes out the essential detection elements (optical flow characteristics) while discarding the computationally intensive pattern recognition process, thereby maintaining detection reliability with reduced computational power requirements
Solution Approach 2:
The detection process is segmented into distinct steps: extracting characteristic features, computing optical flow vectors, and evaluating plausibility criteria. This segmentation allows the system to focus computational resources only on the essential motion analysis rather than performing comprehensive pattern recognition, resolving the contradiction between reliability and computational power
2Measurement precision
If stereo cameras are used for pedestrian detection, then detection precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent makes a standard single camera perform multiple functions by utilizing temporal sequence analysis and optical flow computation. The same camera hardware that captures images for general purposes is also used for precise pedestrian detection through motion analysis, eliminating the need for specialized stereo camera systems while maintaining detection precision
Solution Approach 2:
The patent changes the analysis parameters from spatial stereo vision to temporal motion analysis. By analyzing optical flow vectors across time sequences from a single camera, the system achieves detection precision comparable to stereo cameras but with significantly reduced device complexity and cost
3Measurement precision
If optical flow vectors are computed for all characteristic features, then motion detection accuracy is improved, but computational effort increases
Solution Approach 1:
The patent applies partial action by computing optical flow vectors only for characteristic features that meet specific plausibility criteria rather than for all features. It performs the necessary computational work on a subset of relevant features, maintaining motion detection accuracy while improving processing speed by avoiding unnecessary computations on irrelevant features
Data Source
AI summary
The invention relates to a method for detecting a pedestrian (27) moving in an environmental region of a motor vehicle relatively to the motor vehicle based on a temporal sequence of images (18) of the environmental region, which are provided by means of a camera of the motor vehicle (1), wherein characteristic features are extracted from the images (18) and a plurality of optical flow vectors is determined to the characteristic features of at least two consecutively captured images of the sequence by means of an image processing device of the motor vehicle, which indicate a movement of the respective characteristic features over the sequence, wherein for detecting the pedestrian (27), several confidence metrics are determined based on the characteristic features and the optical flow vectors, and based on the confidence metrics, it is examined if a preset plausibility check criterion required for the detection of the pedestrian (27) is satisfied, wherein the pedestrian (27) is supposed to be detected if the plausibility check criterion with respect to the confidence metrics is satisfied.