Object Tracking Apparatus Using Position Prediction and Similarity Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for object tracking in automotive vehicles using stereo cameras face challenges due to ambient illumination changes, object movement, and similar appearances of pedestrians or vehicles, leading to inaccurate identification of objects across frames.
Innovation Solution
An information processing apparatus that predicts the position of objects based on previous frames and extracts matching objects in current frames using similarity analysis, ensuring accurate tracking by associating positions in the longitudinal, lateral, and depth directions over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If object tracking is performed using conventional techniques based on parallax images, then tracking speed is improved, but tracking accuracy deteriorates due to errors from ambient illumination changes, object movement, and similar object appearances
Solution Approach 1:
The patent introduces an intermediary process that uses multiple detection methods (color information, shape information, position information) to verify and cross-validate object identity across frames. This intermediary verification step mediates between the fast but inaccurate conventional tracking and the need for high accuracy, resolving the contradiction by adding a validation layer that doesn't significantly slow down the overall process.
Solution Approach 2:
The patent changes the parameters used for object identification by incorporating multiple types of information (color, shape, position) instead of relying on a single parameter. By changing from single-parameter tracking to multi-parameter verification, the system maintains speed while improving accuracy through comprehensive parameter comparison.
2Device complexity
If simple object tracking based on previous frame detection is used, then processing complexity is reduced, but reliability deteriorates due to mistaken object identification
Solution Approach 1:
The patent segments the object verification process into multiple independent components: color information extraction, shape information extraction, position information extraction, and similarity comparison. By segmenting the verification process, the system reduces overall complexity through modular design while improving reliability through comprehensive multi-faceted verification of each object.
Solution Approach 2:
The patent performs preliminary extraction and storage of color, shape, and position information from detected objects before the actual tracking and verification process. This preliminary action prepares verification data in advance, reducing the complexity of real-time verification while ensuring reliable object identification through pre-computed comparison parameters.
3Loss of time
If conventional tracking methods are used without considering multiple object attributes, then processing time is reduced, but measurement precision deteriorates due to inability to distinguish similar objects
Solution Approach 1:
The patent applies partial verification by selectively comparing only the necessary attributes (color, shape, position) that are sufficient to distinguish objects in the current scene. By performing partial verification on key attributes rather than exhaustive analysis of all possible object characteristics, the system reduces processing time while maintaining sufficient precision for accurate object distinction.
Data Source
AI summary
An information processing apparatus includes an obtaining unit configured to obtain information in which positions of an object in a vertical direction, positions in a horizontal direction, and positions in a depth direction at multiple points in time are associated with each other; a prediction unit configured to, based on a position of a predetermined object in the information previously obtained by the obtaining unit, predict a position of the predetermined object in the information currently obtained by the obtaining unit; and an extraction unit configured to extract, from the current information, multiple objects that satisfy a predetermined condition corresponding to the position of the predetermined object, and extract, based on a degree of similarity between each image of the multiple objects and an image of the predetermined object, a same object, of the multiple objects in the current information, as the predetermined object in the previous information.