Vehicle Camera Attention Weighting for Pedestrian Crossing Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated vehicle braking systems face challenges in predicting pedestrian crossing intentions accurately, particularly with wide field of view images that include irrelevant information, which can lead to collision risks if pedestrians cross the road while the vehicle is driving.
Innovation Solution
A vehicle control system that uses a machine learning model, including visual transformer layers and a multilayer perceptron, to extract features from vehicle camera images, assign attention weights to regions of interest, and generate a crossing intention prediction output, triggering automatic braking if the prediction exceeds a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If wide field of view camera images are used to capture more traffic information, then the coverage area increases, but the amount of irrelevant information increases reducing prediction accuracy
Solution Approach 1:
The patent divides the wide field of view image into multiple regions and applies different attention weights to different regions. The attention mechanism segments the image processing task, allowing the system to focus computational resources on relevant areas (where pedestrians are detected) while downweighting irrelevant regions, thus maintaining prediction accuracy despite wide coverage.
Solution Approach 2:
The patent implements local quality by assigning different attention weights to different spatial regions of the image based on detected objects. Regions containing pedestrians receive higher attention weights while other regions receive lower weights, allowing the system to maintain high prediction accuracy in critical areas while efficiently processing the entire wide field of view.
2Reliability
If attention weights are assigned to all regions of the image, then the processing completeness improves, but the computational complexity increases
Solution Approach 1:
The patent applies partial action by assigning significant attention weights only to regions where pedestrians are detected, while using lower weights for other regions. This allows the system to maintain processing completeness for the entire image while effectively focusing computational resources on the most critical regions, reducing overall computational complexity.
3Loss of information
If machine learning model processing is applied to the entire image, then the feature extraction completeness improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary object detection to identify pedestrian locations before applying the full machine learning model for crossing intention prediction. This preliminary action allows the system to pre-identify regions of interest, so that subsequent feature extraction and prediction can be focused on these specific areas, maintaining completeness while reducing processing time.
Data Source
AI summary
A method for controlling automated vehicle acceleration and braking includes obtaining an image using at least one vehicle camera of a host vehicle, extracting machine learning model feature inputs based on the obtained image, detecting one or more objects in the obtained image, the one or more objects including at least one pedestrian, assigning attention weights to regions of the obtained image according to locations of the one or more objects in the obtained image, combining the attention weights with corresponding ones of the machine learning model feature inputs according to the regions of the obtained image, executing a machine learning model to generate a crossing intention prediction output associated with the at least one pedestrian, and in response to the crossing intention prediction output exceeding a crossing intention threshold, controlling automatic braking of the host vehicle according to a location of the at least one pedestrian.