Depth estimation with tilted optics for automotive

2018 Fiscal Year Final Research Report

• Project/Area Number: 15K00365
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Intelligent robotics
• Research Institution: Fukuyama University
• Principal Investigator: IKEOKA Hiroshi 福山大学, 工学部, 准教授 (20579966)
• Research Collaborator: HAMAMOTO takayuki
• Project Period (FY): 2015-04-01 – 2019-03-31
• Keywords: 距離推定 / アオリ光学系 / 車載

Outline of Final Research Achievements

The purpose of the study for our depth estimation method with tilted optics is obtaining depth for the circumference of a vehicle. To reduce optical distortion which occurred the error of sharpness value from image distortion, we developed a method which reduce the influence of contrast difference with ground-in-focus imaging. Moreover, we proposed using a neural network to solve the problem about the difference between the optical theory and the real optics for wide view angle of a lens. Additionally, we proposed using a neural network to obtain sharpness values. These proposals have realized a depth estimation method which have wide detectable range with tilted optics.

Free Research Field

コンピュータビジョン

Academic Significance and Societal Importance of the Research Achievements

本研究の主な成果は，距離推定精度の向上およびそれに伴うより推定範囲の広範囲化である．事故予防および自動運転を実現するうえで，車載の全周囲にわたって各物体までの距離を監視することが不可欠である．しかし，全周にわたって距離を推定するには，カメラなどの多くのセンサ・デバイスが必要になるが，デバイス一台で広範囲をカバーした距離推定が可能になれば，車両に搭載するデバイス数を低減でき，コスト面およびデザイン面で非常に有益である．よって，これを安価なアオリ光学系で実現できれば，その研究意義は非常に大きい．