非平衡宽覆盖像面扫描系统的神经网络控制

为扩大遥感仪器像面覆盖宽度,从现有的CCD拼接方法入手,提出了凸轮驱动的动态扫描拼接方法。将电机与凸轮同轴安装,两者做等速旋转运动,带动4片线阵TDI CCD在像面上做往复直线运动,相邻线阵CCD在扫描方向上的成像区域保持一定重叠率,从而实现了动态扫描拼接。分析了由凸轮结构的特殊性造成的系统负载力矩的非平衡特性进行了分析,针对采用常规稳速方法时,凸轮转速在负载变化阶段产生较强波动的情况,提出了常规稳速控制和神经网络相结合的自适应控制方法,并进行了实验分析。分析表明:作用于凸轮轴上的负载力矩与相机位角、凸轮转速成正比。与常规稳速方法相比,应用神经网络自适应控制方法后,系统稳速精度提高了41%,非平衡负载引起的速度波动降低了20%,满足工程需要。

Neural network control to image scanning system with nonequilibrium load and wide coverage

To extend the imaging coverage of a remote sensing instrument,a dynamic scanning assembly actuated by a cam is proposed based on the existing CCD assembly methods.A motor and a cam were coaxially installed and were rotated at a constant speed. Four pieces of linear TDI CCDs were driven by the cam to reciprocate on the focal plane. The imaging areas of the two neighbor piece of CCDs in the scanning direction were overlapped to realize the scanning assembly. The nonequilibrium load feature of the system caused by the special structure of the cam was analyzed in detail. In consideration of that the cam speed would be fluctuated greatly due to the load variation by using the conventional speed control method,a neural network controller combined with the conventional speed control method was put forward. The proposed method was analyzed,and the analysis indicates that the load torque on the cam axis is directly proportional to the camera azimuth and cam speed,respectively. Furthermore, compared with that of conventional speed control method,the speed precision is enhanced by 41% when the neural network controller is used,meanwhile,the speed-fluctuation is reduced by 20%. The whole design could satisfy project requirements.