Camera Tilt Detection Using Image Line Extraction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional techniques for detecting near miss cases by analyzing image data from vehicle-mounted cameras are compromised if the image capturing direction is tilted, leading to invalid analysis results.

Innovation Solution

An image processing device that acquires and processes image data to extract straight lines using the Canny filter and Hough transform, detecting tilt by analyzing the angle between these lines and the vertical axis, thereby determining if the image capturing direction is abnormal.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If image data is captured by a vehicle-mounted camera for near miss detection, then near miss cases can be analyzed, but the analysis becomes invalid when the camera is tilted

Engineering Contradiction:

Improvevalidity of near miss analysisVSAvoidaccuracy of image capturing direction

Core Design Contradiction:

ReliabilityVSMeasurement precision

Solution Approach 1:

The system performs preliminary detection of camera tilt angle before conducting near miss analysis. By detecting the vertical and horizontal angles of the image capturing direction in advance, the system can determine whether the camera is properly positioned and exclude tilted images from analysis, ensuring the reliability of subsequent near miss detection

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces an intermediary detection mechanism that measures the camera's tilting state as a separate step. This intermediary measurement of vertical and horizontal angles acts as a gatekeeper, filtering out improperly positioned cameras before they can compromise the near miss analysis process

Inventive Principle:

Principle #24Intermediary (Mediator)

2Quantity of substance

If the entire image area is used for analysis, then more data is available, but peripheral distortion and noise reduce measurement accuracy

Engineering Contradiction:

Improveamount of image dataVSAvoidaccuracy of line angle extraction

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The system segments the image area into a central analysis region and peripheral exclusion zones. By extracting straight lines only from the central portion of the image while excluding the periphery where distortion occurs, the system maintains measurement precision while still obtaining sufficient data for accurate tilt detection

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality control by assigning different processing treatments to different regions of the image. The central region is used for precise line extraction and tilt measurement, while the peripheral regions are excluded or given lower weight, ensuring that high-quality data drives the measurement process

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS10230946B2Method of detecting tilt of image capturing direction and apparatus of detecting tilt of image capturing direction

Publication Date: 2019.03.12 FUJITSU LTD

AI summary

An image processing device includes an acquisition unit, an extraction unit, and a detection unit. The acquisition unit acquires image data for each of frame numbers from a drive recorder or the like. The extraction unit extracts a plurality of straight lines from the image data on the basis of drive recorder information. The detection unit detects a tilt of an image capturing direction of a camera according to a magnitude of an angle between a vertical axis direction of the image data and the extracted plurality of lines.