Pedestrian Detection via Velocity Change Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing pedestrian detection systems using optical flow struggle with accuracy, especially when the space model needs to be assumed in a short time, leading to inaccurate detection of moving objects, particularly pedestrians, due to their slow and irregular movements.
Innovation Solution
A pedestrian detection system that includes an image pickup unit, a controller with a characteristic point extractor, movement information calculator, coordinate converter, object attribute determinator, and pedestrian determinator, which processes images to identify characteristic points, calculates movement velocities, and determines if objects are pedestrians based on velocity changes in an overhead view coordinate system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If optical flow of background is calculated based on space model and host vehicle movement, then moving objects can be detected, but detection accuracy deteriorates when space model assumption time is short
Solution Approach 1:
The patent segments the detection process into two independent parts: (1) detecting moving objects by comparing optical flow magnitude thresholds, and (2) detecting pedestrians by analyzing velocity changes of detected moving objects. This segmentation eliminates the need for inaccurate space model assumptions while maintaining detection capability.
Solution Approach 2:
The patent changes the detection parameter from optical flow magnitude (which requires space model) to velocity change over time (which does not require space model). By calculating the difference in optical flow values between consecutive frames and comparing against thresholds, the system achieves accurate detection without relying on host vehicle movement models.
2Ease of operation
If optical flow method is used to detect moving objects, then detection can be performed, but pedestrian detection accuracy deteriorates due to slow and irregular pedestrian movement
Solution Approach 1:
The patent first detects all moving objects using optical flow magnitude comparison, then performs a second stage analysis on the detected moving objects by calculating their velocity changes over time. This preliminary detection followed by refined analysis allows the system to identify pedestrians even when their movement is slow or irregular, improving overall detection accuracy.
Solution Approach 2:
The patent uses feedback by continuously calculating velocity changes of detected moving objects across multiple frames. By monitoring the temporal变化 of optical flow values and comparing against predetermined thresholds, the system can identify pedestrians based on their characteristic velocity patterns, even when movements are subtle or irregular.
Data Source
AI summary
A pedestrian detecting method including: picking up an image of a scene in front of a vehicle; extracting characteristic points from the picked up image; calculating a velocity at which, and a direction in which, each of pixels representing the respective characteristic points moves in the image; expressing coordinates of each of the pixels in a coordinate system corresponding to an overhead view; determining whether the characteristic points represent a two-dimensional object or a three-dimensional object, and if it is determined that they represent the three-dimensional object, whether or not the three-dimensional object is a moving object, on the basis of the calculated velocity and direction of each of the pixels and the coordinates thereof in the overhead view; and determining whether or not the moving object is a pedestrian on the basis of change in the velocity at which the moving object moves in the image.