Vehicle Behavior Inference Using Excluded Video Movement Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle behavior inference and unsafe driving detection systems require connection to the in-vehicle network to function, limiting their ability to infer vehicle behavior independently.
Innovation Solution
A system that calculates movement vectors from video images, infers areas indicating movable objects, excludes these vectors, and infers vehicle behavior without network connection, using surrounding-area and driver posture information for unsafe driving detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the apparatus is connected to the in-vehicle network to acquire vehicle information, then the vehicle behavior inference can be performed using vehicle data, but the apparatus cannot function independently when not connected to the in-vehicle network
Solution Approach 1:
The patent introduces a camera as an intermediary device to capture images of the vehicle's front area, replacing the need for direct connection to the in-vehicle network. The camera system acts as a mediator that provides visual data about the vehicle's movement and surroundings, enabling behavior inference without relying on vehicle-mounted sensors or network connectivity.
Solution Approach 2:
The patent substitutes the mechanical/electronic sensor system (vehicle speed sensors, steering angle sensors, yaw-rate sensors) with an optical imaging system (camera). By replacing the traditional vehicle data acquisition mechanism with image capture and processing, the system achieves independence from the in-vehicle network while maintaining the ability to infer vehicle behavior.
2Productivity
If movement vectors from all areas are used for behavior inference, then the inference can be performed, but the accuracy is reduced due to inclusion of movable objects like pedestrians and other vehicles
Solution Approach 1:
The patent segments the image area into different regions, identifying and separating the area occupied by movable objects (pedestrians, other vehicles) from the area occupied by the vehicle itself. By dividing the image into these distinct segments, the system can selectively process only the relevant vehicle area while excluding misleading movement vectors from other objects.
Solution Approach 2:
The patent extracts and removes movement vectors corresponding to movable objects (pedestrians, other vehicles) from the overall movement vector data. By taking out these extraneous vectors that do not represent the vehicle's actual movement, the system maintains processing efficiency while significantly improving the accuracy of vehicle behavior inference.
Data Source
AI summary
A movement vector calculation unit calculates movement vectors between frames of a video image of an area ahead of a vehicle. The video image is input as a moving image. An area inference unit infers an area indicating a movable object included in the video image of the area ahead of the vehicle. A vector excluding unit excludes, from among the calculated movement vectors, movement vectors in an area inferred as being the area indicating the movable object. A behavior inference unit infers behavior of the vehicle based on the movement vectors from which those in the area inferred as being the area indicating the movable object have been excluded.