Monitoring Technology Of Dead Pine Wood With Pine Wood Nematode Disease Based On Aerial Image Of Drone

Pine tree is one of the important components of China’s forest resources,and it is widely distributed in artificial forest areas and nature reserves.Pine wood nematode disease has a very strong infectivity and spread,also known as "cancer of pine trees".It is mainly transmitted by pine brown Monochamus alternatus as intermediate host and gradually erodes the healthy pine wood around the diseased pine wood.The infected pine will gradually lose water and discolor the needles,and eventually the entire plant will die completely.Since its discovery,pine wood nematode disease has caused outbreaks in many provinces in China,causing huge economic losses.Due to the vastness of the forest area,common artificial field surveys have problems of visual blind spots and time-consuming and laborintensive problems,and it is difficult to perform timely and effective monitoring of forest area diseases.The gradual popularization of UAV remote sensing technology has made up for the problems of poor efficiency and time-consuming labor of traditional monitoring methods,Combined with remote sensing information can also accurately locate disease pines,providing a more convenient technical means for disease monitoring in forest areas.In response to the above situation,this article uses commercial small UAV as flight platforms and combines image processing technology to monitor diseased pine trees in the forest area.Taking the pine wood nematode outbreak area in Jian’ou City,Fujian Province as the investigation and test area,the remotely-sensed shooting and postprocessing of the data were performed on the ward,and the deep-learning framework was used to realize the automatic recognition of the diseased pine wood images in the test area.Analysis and discrimination of disease conditions provide accurate data for disease prevention and monitoring.The work of this paper mainly includes:(1): With small commercial unmanned aerial vehicle(UAV)as the flight platform,continuous image collection of designated disease outbreak locations is performed every 20 days,providing reliable data support for sample training in the later period.(2): The images taken by the UAV are pre-processed by screening,cropping,and other methods to improve the quality of the proofs.The multi-channel threshold segmentation method using color features as the means and the ultra-red color feature combined with the maximum between-class variance method were used to discriminate the dead pines in remote sensing images.The results show that the multi-channel threshold method has higher recognition accuracy,but the previous work is relatively time-consuming.Both methods have the disadvantages of missed judgment,misjudgment of the disease pine and the inability of the algorithm to be widely used in the treatment process.There is also a gap with the accuracy of manual identification,and it is impossible to achieve efficient and accurate discrimination of forest diseases.(3): Marking and collating the remote sensing data in the early stage to make a data set.When making the data set,the disease pine was divided into level 1,2,and 3(corresponding to the early stage of pine infection,severe infection,and death from infection),has achieved a more accurate discrimination of pine disease in the ward.(4): Using the Faster R-CNN,YOLOv3,and SSD object detection frameworks in deep learning field to identify ward samples,comparing the processing results of each framework,it is found that the Faster R-CNN framework has the best effect,and the recognition accuracy rate exceeds90 %.This method not only overcomes the situation of misjudgment and omission in the traditional recognition methods,but also accurately judges the disease level of the pine wood in remote sensing image,and realizes accurate and efficient automatic recognition.