An automatic and precise calculation method for the pose and pose of rotating linear array scanning images

Abstract

The invention proposes a precise calculation method for the pose of a multi-view rotating linear array image based on an equivalent frame image. The method firstly assumes that the linear scan camera captures the image under ideal placement conditions, and projects it onto the tangent plane through tangent projection to generate an equivalent frame image with the same internal and external parameters as the rotated scan image. Then, the multi-view motion structure recovery is performed according to the image, and the highly reliable initial values ​​of the image pose and scene landmarks are obtained. Finally, according to the strict imaging model of rotational scanning imaging, the pose parameters of the above-mentioned equivalent frame image and the coordinates of landmark points are used as the initial values ​​of the beam method adjustment, and the precise pose parameters of the rotational scanning line array image are obtained by iterative optimization.

Description

technical field [0001] The invention relates to the field of image geometry processing and three-dimensional reconstruction, in particular to an imaging model of a rotating linear array scan image and a pose automatic precision solution method. Background technique [0002] The hyperspectral imager in the line scan imaging mode, with its good spatial-spectral consistency, the best spatial-spectral resolution, and good coordination with various imaging platforms (UAVs, manned aircraft, satellites, etc.) , is currently the mainstream hyperspectral camera imaging method. This type of camera can only acquire one scan line image for one exposure. In order to obtain continuous scene images, the camera is usually fixed on a rotating pan-tilt, and the pan-tilt drives the camera to perform uniform circular motion to acquire another dimensional spatial data. [0003] For the hyperspectral image collected by the whole scene line scan, its imaging does not conform to the pinhole imagin...

AI Technical Summary

Problems solved by technology

[0004] However, the form of the model is relatively complex, and it is impossible to directly use the model to obtain the analytical solution of the relative position relationship between images. It is necessary to use the optimal estimation method (such as the Levenberg-Marquardt method) for iterative optimization, and finally obtain the least squares solution of all parameters

In this algorithm, the accuracy of the initial value of the parameter has a great influence on the correctness of the result. A good initial value of the parameter can make the algorithm quickly converge to the optimal solution, while a bad initial value will cause the algorithm to converge slowly and to a local optimal solution even fails to converge

Images