Pedestrian Separation via Image Signature Spatial Relationships
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Neural networks trained as regressors fail to effectively perform crowd separation, which is crucial for predicting pedestrian movement and ensuring safe autonomous or semi-autonomous driving in crowded environments.
Innovation Solution
A method involving supervised machine learning to generate and process image signatures of pedestrians, detecting unique combinations of identifiers indicative of spatial relationships between regions and bounding boxes, and using these to locate and distinguish pedestrians based on spatial relationships learned from densely positioned pedestrian test images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If neural networks trained as regressors are used for crowd separation, then the system can process crowd images, but it fails to achieve accurate pedestrian identification and separation
Solution Approach 1:
The patent divides the crowd image processing into distinct stages: generating image signatures, detecting unique identifier combinations, and locating individual pedestrians. This segmentation allows each stage to be optimized independently, with the signature generation focusing on regional features and the detection stage focusing on unique combinations, thereby resolving the contradiction between processing capability and separation accuracy.
Solution Approach 2:
The patent performs preliminary action by generating image signatures and detecting unique identifier combinations before final pedestrian location and identification. This preliminary processing of spatial relationships and identifier patterns enables the system to pre-filter and pre-organize data, making the final separation and identification stages more accurate and reliable.
2Device complexity
If traditional regressor methods are used, then the system structure is simple, but it fails to detect unique spatial relationships in densely positioned pedestrians
Solution Approach 1:
The patent introduces image signatures as an intermediary representation that captures spatial relationships and identifier combinations. This intermediary structure bridges the gap between simple image input and complex pedestrian separation, allowing the system to detect spatial relationships without requiring overly complex network architectures. The signatures serve as a mediator that encodes spatial information in a structured, analyzable format.
Solution Approach 2:
The patent transforms the problem from direct spatial coordinate detection to identifier combination detection within image signatures. By introducing the dimension of identifier combinations and their spatial relationships, the system can detect pedestrian positions and relationships without directly measuring complex spatial coordinates, thereby simplifying the detection process while improving accuracy.
3Quantity of substance
If the system processes all identifier combinations, then comprehensive coverage is achieved, but computational efficiency decreases due to processing irrelevant combinations
Solution Approach 1:
The patent applies local quality by focusing computational resources on detecting unique identifier combinations that are locally significant for pedestrian separation. Rather than uniformly processing all identifier combinations, the system identifies and processes only those combinations that provide meaningful spatial relationship information, thereby maintaining comprehensive coverage of relevant patterns while improving processing efficiency by eliminating redundant computations.
Data Source
AI summary
A method for crowd separation, the method may include receiving or generating a signature of an image of a first plurality of pedestrians, wherein the image comprises regions, wherein the signature of the image comprises descriptors of the regions, wherein each descriptor of a region is associated with a region and comprises a set of identifiers that identify content included in the region and in a vicinity of the region; detecting, within the descriptors of the regions, unique combinations of identifiers that are indicative of spatial relationships between the regions and bounding boxes that surround pedestrians of the first plurality of pedestrians; wherein the unique combinations are learnt during a supervised machine learning process that is fed with test images of densely positioned pedestrians; and locating the pedestrians of the first plurality of pedestrians based, at least in part, on the spatial relationships related to detected unique combinations and to locations of the regions.