The Geometric Transformation Model of Two Views Based on the Line-Scan Camera Imaging Model

Abstract

AbstractWith the increasing use of line-scan camera, it is difficult to ensure that its sensor is parallel to the target in imaging. This will make the target present a different appearance in linear array images. When two linear array images are registered, it is crucial to choose a suitable geometric transformation model. However, the imaging model of line-scan camera is different from that of frame camera, and the classical geometric transformation model of frame image, namely perspective transformation model, does not conform to geometric transformation of linear array image. Therefore, according to the imaging model of line-scan camera, the geometric transformation model of linear array image is derived in this paper. To obtain linear array images, an acquisition system is built. The geometric transformation model established in this paper and perspective transformation model are used to register the linear array images respectively. The registration results based on the geometric transformation model of linear array image show that the two linear array images can be completely aligned and the root mean square error of feature points is smaller. On the contrary, the registration results based on perspective transformation model show that the two images are not alignment, and the root mean square error of feature points is larger, which also indicates that it is not suitable for geometric transformation of linear array images.KeywordsLine-scan cameraImaging modelGeometric transformation modelLinear array image