Image guidance technology of rocket sub-stage landing based on strapdown inertial navigation

In view of the full ballistic flight characteristics of the launch vehicle sub-stage, a strapdown inertial navigation model in the launch inertial system is established for the rocket sub-stage and the strapdown solution is realized in this coordinate system. In order to improve the reliability and accuracy of the autonomous landing of the rocket sub-stage under the shaking state, the guidance technology based on image recognition is introduced at the end of the landing. For the task of detecting and tracking multi-scale cooperative targets in complex backgrounds, we propose a target detection model based on the balanced attention feature pyramid and an improved DeepSort algorithm on the basis of the acceleration motion model, which have achieved the accurate detection and tracking of cooperative targets under the condition of shaking the rocket.