No abstract

The alignment accuracy and stability of shipborne electro-optical system are affected by external attitude disturbance and internal non-linear disturbance. Therefore, this study examined a new digital sliding mode control based on a novel reaching law with an n-order disturbance compensator to solve this problem. Specially, the proposed novel reaching law was a chatter-free method, in which, as the switching gain was replaced by an adaptive function, chattering on the sliding surface tended to zero. For perturbed systems, the new sliding mode controller was able to guarantee a narrow quasi-sliding-mode domain and the ultimate width was the order of O(T 3 ), such that the narrower the quasi-sliding-mode domain width was, the more robust toward nonlinear disturbances. Both mathematical calculations and simulations verified the convergence and stability of the new controller. Finally, based on a compound control strategy that combined feedforward control with a sliding mode control position loop and proportional-integral velocity loop control structure, shipborne equivalent dynamic target tracking experiments were performed. The results, compared with existing controllers, showed that the robustness of the new controller to nonlinear disturbances was stronger and the tracking system accuracy clearly improved, while the influence of sliding mode chattering on the system was avoided.INDEX TERMS Shipborne electro-optical system, nonlinear disturbance, digital sliding mode control, chattering, reaching law, disturbance compensator, chattering-free, quasi-sliding-mode domain.

Image correction technology is very important both for image mosaic and for measurement. To overcome the defects and limitations of polynomial and TPS models, basing on polynomial and TPS correction models, an improved algorithm is proposed in this paper. In this method, the processing is divided into two stages, global correction and local correction, and different screening methods of GCPs are adopted in different stages. By fully utilizing all GCPs, the respective correction advantages of the two sub-models can be efficiently demonstrated, and, in the meanwhile, the ultra-plus effect of this combination can be stimulated. Eventually, image correction accurcy can be improved. Simulation results show that, comparing with currently available approximate geometric correction methods, this method has high accuracy, good robustness and can effectively coordinate the conflict of the global correction and local correction.

In order to obtain the scale information of ship targets effectively, we proposed an improved Faster R-CNN algorithm which integrated multi-scale region proposal and pooling of ROI, visual attention mechanism and rotation region regression and suppression, and ship targets can be positioned by the rotate quadrangle bounding boxes to obtain the scale information of them. Our improved model is based on the standard Faster R-CNN and is maintained through end-to-end training.

No abstract