Real-time compensation method based on motion prediction of least squares support vector machine (LS-SVM)

Abstract

The invention discloses a real-time compensation method based on the motion prediction of a least squares support vector machine (LS-SVM) in a stable imaging system. The method comprises the followingsteps: firstly, establishing an LS-SVM motion prediction model and setting a parameter; then collecting motion detection data of the imaging system so as to train the motion prediction model; then, outputting a motion vector predicted value of a certain time in the future when a motion vector value of the current time is input; and finally, carrying out digital-to-analogue conversion on the motion vector predicted values by a stable compensation system to generate a compensation vector so as to control a compensation mechanism to carry out real-time compensation on a current motion of a camera. The invention can quickly and accurately predict the current motion state of the camera under the condition of an unknown camera motion model so as to achieve the purposes of detecting, delaying, predicting and compensating the motions, realize the real-time compensation of the motions of the camera, and improve the working performance of the stable imaging system.

Description

technical field [0001] The invention relates to the technical field of computer imaging, in particular to a real-time compensation method based on least square support vector machine motion prediction in a stable imaging system. Background technique [0002] Digital imaging technology has been widely used in aerospace / aviation remote sensing, military reconnaissance, civilian photography and other fields. During the imaging process of the camera device, due to the trembling and vibration of the camera device carrier, such as satellites, aircraft, vehicles, ships, etc., or the hand shake of the photographer, the image plane shakes during the imaging integration time of the image sensor, which in turn leads to the acquisition of blurred images, reduced resolution, and degraded image quality. [0003] Active stabilized imaging is a technology that uses mechanical, optical, or digital processing methods to actively compensate for imaging carrier motion or image plane shake thro...

AI Technical Summary

Problems solved by technology

However, the generalization ability of the BP neural network method is poor, and it cannot meet the requirements of the stable imaging system for prediction accuracy and stability; its operation speed is also slow, and it is difficult to meet the requirements of real-time prediction output; in addition, the network initial structure parameters ( Mainly the number of neurons in the hidden layer) is difficult to determine, and the operability is poor

Images