Research On The Technology Of Micro Unmanned Indoor Target Search And Approach

Micro UAV generally refers to unmanned aerial vehicle smaller than 10 cm in size,of which the load capacity is not large enough to carry high-performance sensors and onboard computer.Therefore,there are difficulties in performing autonomous flight missions for Micro-UAV.In this paper,an indoor target person searching scheme is proposed in navigation and recognition by assembling a monocular camera and laser sensors on a Micro-UAV.In order to search the target person without missing in the complex indoor environment with multi rooms,this paper realizes an “along the edge” searching strategy of micro UAV Based on the path planning of laser ranging,and suppresses the tremor caused by the laser data noise by introducing the auto disturbance rejection control in the yaw angle,and realizes the ergodic exploration of the main indoor areas.In order to improve the problem of poor terrain trafficability and redundant search path,a monocular vision navigation algorithm based on ground station assisted cloud computing is designed.The algorithm uses full convolution neural network to restore the image depth of the forward-looking image captured by the visual sensor,and then uses the depth threshold graph and image gradient analysis to plan the path,so as to complete the navigation task when the laser cannot work normally and optimize the redundant search path in the process of extending the flight.In the approach scheme,we first design a joint image detection and tracking target recognition scheme.In the aspect of target detection algorithm,it reduces the calculation time consumption of the algorithm by data set processing and parameter pruning optimization of the existing Yolo neural network.At the same time,by associating the face,hand and torso in the detection process,the tracking of the target can be maintained when some features of the target are occluded.Then,this paper introduces the KCF image tracking algorithm to train the filter with the result of Yolo detection,so as to further improve the recognition speed without affecting the accuracy.Finally,the proportional guidance method is used to generate the control instructions in the process of the UAV approaching the target.