Research On Target Detection And Location Of Photovoltaic Clean Aircraft

In photovoltaic power generation systems,photovoltaic modules are usually erected in highlands or remote areas.Sand and dust or other particles in harsh environments will adhere to their surfaces,which will reduce the efficiency of photovoltaic power generation and generate hot spot effect,bringing economic losses to the photovoltaic power station.This paper proposes to use the aircraft equipped with a camera and a cleaning device to complete the cleaning task efficiently.At the same time,aiming at the key problem that the aircraft's GPS positioning error is large and it can't reach the specified location accurately.Combined with deep learning target detection algorithm to carry out terminal target recognition of photovoltaic modules and guide aircraft positioning.The main work of this paper is as follows:(1)In view of the slow recognition speed and poor recognition effect of traditional image processing methods,nearly 500 VOC data sets of photovoltaic modules were produced and expand the size of the data set through data augmentation.On the TensorFlow deep learning framework,one-stage detection method SSD and two-stage detection method Faster R-CNN are compared,and the SSD detection algorithm with faster detection speed is selected as the design algorithm.(2)In view of the fact that the SSD algorithm under the photovoltaic module detection experiment has a poor effect on small targets and misses detection.it is proposed to replace VGG16 with ResNet and combine it with deconvolution to obtain the DSSD algorithm;At the same time,according to the characteristics of this design,a large amount of computational redundancy is generated in the frame generated by the SSD network,so the network is cut to reduce the amount of calculation and speed up the operation;Finally for the actual situation,the camera movement process Blur,using Wiener filtering and regularization deconvolution to get a clear image.(3)Using the DJI aircraft Onboard SDK and the optimized convolutional neural network optimization algorithm to design a photovoltaic cleaning system.At the same time,the photovoltaic cleaning device was designed.Under the experimental conditions,it was proved that the aircraft can complete the tasks of photovoltaic component identification and photovoltaic component cleaning.