Monitoring Of Blooming Mikania Micrantha Based On Uav Remote Sensing

Mikania micrantha is one of the ten most harmful weeds in the world.It has been widely distributed in Guangdong,Guangxi and other regions of my country,posing a great threat to the ecosystem.How to improve the monitoring accuracy of M.micrantha and provide high-quality spatial distribution information for the control of M.micrantha is one of the key means to prevent M.micrantha.The existing M.micrantha monitoring technology is mainly based on manual field monitoring and satellite remote sensing monitoring,but there are deficiencies such as high cost and low accuracy.Unmanned Aerial Vehicle(UAV)remote sensing has the advantages of low cost,strong timeliness and high spatial resolution,but the monitoring of M.micrantha by UAV has not been reported at home and abroad.Therefore,this project takes UAV as the monitoring carrier,takes the blooming M.micrantha under complex environment as the research object,and takes the monitoring mechanism of precise identification,monitoring and risk assessment management of M.micrantha as the research target to carry out the monitoring of invasive alien plant M.micrantha based on UAV remote sensing.The research contents are as follows:(1)Machine learning algorithm and color space algorithm were used to identify blooming M.micrantha.In order to solve the problem of low identification of blooming M.micrantha under complex environment in the wild,Otsu-K-means algorithm,RGB and HSV color space algorithms were used.The results show that Otsu-K-means algorithm has a high precision rate of M.micrantha,which can reach more than 80%,but it will lose some image information during image segmentation,which leads to the unsatisfactory recall rate.When RGB and HSV color space algorithms are used,the best recognition effect is to set the value range of each channel at the sampling average value ± 2* standard deviation as the threshold value,with a high recognition recall rate of about 80%.However,RGB color space algorithm is not able to recognize M.micrantha and land,resulting in the accuracy is not ideal.Therefore,the high precision rate of K-means algorithm and the high recall rate of color space algorithm have certain potential of complementary fusion.(2)The K-means-RGB-HSV algorithm,which combines machine learning and color space algorithms,was applied to identify blooming M.micrantha.Aiming at the problems of low recognition recall rate of machine learning algorithm and low precision rate of color space algorithm,this paper proposed the image segmentation based on K-means algorithm,and the K-means-RGB algorithm and K-means-HSV algorithm of color space algorithm were applied on the basis of segmentation results.Experiments prove that the overall performance of the two algorithms is better than before the fusion,realizing the complementary advantages.In order to further improve the recognition effect,K-means-RGB-HSV algorithm based on K-means-RGB algorithm accounting for 48.65% and K-means-HSV algorithm accounting for 51.35% was proposed.Compared with other algorithms,K-means-RGB-HSV algorithm has the best recognition effect for blooming M.micrantha.It can achieve 96.70% recall rate and 85.29% F1-score value.Under the centimeter resolution accuracy,the algorithm realizes the precise monitoring of blooming M.micrantha based on UAV,and solves the shortages of traditional monitoring.(3)Damage grade of blooming M.micrantha was evaluated by Fuzzy Analytic Hierarchy Process(FAHP)based on the identification results.Based on the status of the monitoring area,the flower biomass of M.micrantha corresponding to the unit pixel point of the monitoring area image is 0.14～0.17(g),and the total biomass is 2.98～3.39(g).Based on the FAHP,the comprehensive assessment model of blooming M.micrantha was constructed,which included five main evaluation factors and twelve specific influencing factors,including invasion factor(36%),fitness factor(29%),diffusivity factor(11%),harmfulness factor(19%)and prevention and control feasibility factor(5%).Based on the results of the calculation,2D and 3D visualized heat maps were generated.Four grade IV risk areas with a risk value of 3.3 such as H13 and 12 grade III risk areas with a risk value of 2～2.99 such as G13 were determined.The model can clearly reflect the damage of blooming M.micrantha and accurately locate the invasion high-risk areas.