3D Object Detection Device Irregularity Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional three-dimensional object detection devices face challenges in accurately distinguishing natural objects like plants and snow from man-made objects, particularly in processing large numbers of patterns and preventing mistaken identification of natural objects as vehicles in adjacent traffic lanes.
Innovation Solution
The device calculates an irregularity evaluation value based on differential waveform or edge information from captured images, determining if the value exceeds a threshold to assess whether detected objects are natural or man-made, thereby preventing misidentification of natural objects as vehicles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If pattern matching with multiple shrubbery patterns is performed to detect natural objects, then detection coverage of natural objects is improved, but processing load increases significantly
Solution Approach 1:
The patent extracts the key characteristic of natural objects (irregularity) from the complex pattern matching process. Instead of comparing images against multiple stored patterns of shrubberies and snow, the system calculates an irregularity evaluation value based on differential waveform information or edge information to identify natural objects, significantly reducing processing load while maintaining detection coverage
Solution Approach 2:
The patent changes the detection parameter from pattern similarity matching to irregularity evaluation. By calculating irregularity evaluation values based on differential waveform information or edge information and comparing against a threshold, the system efficiently identifies natural objects without requiring extensive pattern libraries
2Adaptability or versatility
If pattern matching is used to detect all three-dimensional objects, then detection capability is improved, but accuracy in distinguishing natural objects from vehicles deteriorates
Solution Approach 1:
The patent applies different detection methods to different object types. Natural objects are identified using irregularity evaluation based on differential waveform or edge information, while other objects are detected through conventional means. This localized approach to quality assessment improves accuracy in distinguishing natural objects from vehicles
Solution Approach 2:
Instead of trying to identify what objects are present through pattern matching, the patent inverts the approach by identifying what is irregular (natural objects) versus what is regular (man-made objects like vehicles). This inversion improves distinction accuracy between natural objects and vehicles
3Reliability
If the detection threshold is lowered to prevent false positives of natural objects as vehicles, then false positive rate increases, but detection precision of vehicles decreases
Solution Approach 1:
The patent introduces an intermediary irregularity evaluation step between image capture and vehicle detection. By calculating irregularity evaluation values and comparing against a threshold, the system acts as a mediator to filter out natural objects before they can be misidentified as vehicles, maintaining both false positive prevention and vehicle detection precision
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A three-dimensional object detection device comprises three-dimensional object detection units (33, 37) for detecting three-dimensional objects based on image information of the rear of a vehicle from a camera (10), a natural object assessment unit (38) for assessing that a detected three-dimensional object is a natural object (Q1) including plants or snow based on an irregularity evaluation value calculated based on a first pixel number of first pixels representing a first predetermined differential in the differential image containing the detected three-dimensional object and a second pixel number of second pixels corresponding to the three-dimensional object and representing a second predetermined differential greater than the first predetermined differential, and a control unit (39) for controlling the various processes; the control unit (39) suppressing the assessment that the detected three-dimensional object is another vehicle (VX) when the detected three-dimensional object is assessed by the natural object assessment unit (38) to be a natural object (Q1).