Gaussian Processes for Camera Calibration Image Selection

Abstract

Camera calibration is a crucial pre-processing for 3D related computer vision applications. When non-expert calibration operators capture calibration images, they strongly require guidance what kind of images must be taken. When non-expert operators have no guidance, they tend to increase the number of acquired images by assuming a lot of acquired images may contain enough good images. However this strategy has some drawbacks: (1) longer time consumption (optimization computes inverse matrix at every iteration), (2) the more images are used, the less accuracy is increased (inefficient!!), (3) the acquired images may contain some bad images due to motion blur.

This work is to relax the pressure on such non-expert operators during image acquisition. The proposed method asks such operators to take calibration images with variety of position and orientation and then the method takes a subset of good quality images. Thanks to Gaussian Process modeling, the proposed method can select a near optimum subset with some accuracy guarantee in the sense of submodularity.

Publication

Please contact me if you want any pre-prints.

• Yuji Oyamada, Gaussian Processes for Efficient Plane-based Camera Calibration, International Workshop on Frontiers of Computer Vision (IW-FCV), accepted as oral, 2020. [slide, bib]
• 小山田 雄仁, 勝俣 槙太郎, 劣モジュラ関数最大化によるカメラ校正に適した画像選択, 第25回画像センシングシンポジウム, IS2-17, pp. 1-4, Jun., 2019. [poster]