Multi-camera data splicing method based on strict imaging model

Abstract

The invention provides a multi-camera data splicing method based on a strict imaging model. The method comprises steps that an ideal multi-camera splicing focal plane is established according to design parameters of at least two cameras; according to the ideal splicing focal plane and ideal auxiliary parameters, a strict imaging model after multi-camera splicing is established; according to actual parameters, a strict imaging model of each camera is established; according to the strict imaging models of the cameras, a positive and negative computation model between image points and ground points is established; after multi-camera splicing, image points of the cameras corresponding to any image point on an image are calculated, re-sampling is further carried out, after multi-camera splicing, an image after splicing is outputted from the ideal image plane. Through the method, multi-camera splicing precision can be improved, problems of low multi-camera splicing precision and poor stability in the prior art are solved, through experiment verification, splicing precision is stable, splicing is not influenced by an image texture status factor, and the splicing precision is high.

Description

technical field [0001] The invention relates to the field of optical remote sensing satellite camera imaging, in particular to a multi-camera data splicing method based on a strict imaging model. Background technique [0002] In the optical visible light remote sensing satellite imaging process, high resolution and wide field of view are still a pair of contradictions. At present, in order to ensure the acquisition of images with a higher spatial resolution and a certain width, the multi-chip TDICCD splicing method and the multi-camera splicing method are mostly used in the payload design. Among them, the multi-camera stitching process mostly uses the method based on feature point matching to process in the air domain or frequency domain. However, in remote sensing data engineering production, due to the low stability of the internal and external orientation elements of multiple cameras, the stitching accuracy is often not guaranteed. Remote sensing data generally use imag...

AI Technical Summary

Problems solved by technology

The method based on image features extracts features such as image boundaries, contour lines, and inflection points for matching, constructs a system of equations, and obtains transformation parameters through numerical calculation. The disadvantage of this method is that it must rely on image features. If there is an error in the matching between them, the results obtained will also have a large error; secondly, the feature points are very sensitive to noise and occlusion, even if there are good feature points, they are often unable to calculate the alignment due to noise or occlusion

The method based on frequency domain matching uses the cross-correlation function to measure the similarity of two images. This method is simple and intuitive, but it is difficult to deal with the small-angle rotation and small scaling between images. At present, sub-pixel interpolation is more commonly used. method, but also to search the entire image space, computationally expensive and sensitive to image noise

[0004] In view of this, it is necessary to propose a multi-camera data stitching method based on a strict imaging model to achieve multi-camera stable High precision splicing

Images