Research On Wind Response And Collapse Prediction Of Rapeseed At Seedling Stage Based On UAV Image

Oilseed rape is an important oilseed crop in China,and its yield and quality have been reduced due to its collapse,which has seriously restricted the development of oilseed rape cultivation in China.Studies on the effects of different periods of inversion on economic traits and yield of rapeseed have shown that the earlier the period of inversion,the greater the impact of inversion on yield and economic traits.Therefore,it is important to establish the monitoring and prediction of inversion resistance at the seedling stage of oilseed rape.In this paper,the wind response characteristics of rapeseed leaves under the action of underwash airflow were obtained through unmanned aircraft underwash airflow excitation,while static indicators such as rapeseed petiole bending characteristics and rapeseed vegetation index were tested to analyse the intrinsic relationship between the wind response characteristics and the obtained static indicators.The main findings are as follows.The main research contents and results are as follows.(1)Simulation study on the airflow distribution and rape flow-solid coupling of UAV underwash: a three-dimensional model of UAV was established through inverse modelling,and the wind velocity distribution of UAV underwash flow field was simulated based on CFD computational fluid dynamics,and the results showed that when the hovering height was 2m,the coefficient of variation of airflow velocity in the rape canopy distribution area was the smallest and the wind velocity distribution was the most uniform,and the wind velocity size reached 3m/s,which met the test requirements;In addition,through the oilseed rape-UAV downwash airflow fluid-solid coupling analysis of the movement state of the seedling oilseed rape model under the action of downwash airflow,the preliminary determination of the movement trend of oilseed rape,the results show that:the pressure at the leaf tip of oilseed rape leaves is significantly greater than other parts,the vertical direction component of leaf tip displacement is much greater than its horizontal direction component,and the leaf area of oilseed rape and the mechanical properties of the material have a more significant effect on the displacement of the leaf tip.(2)Research on the wind response characteristics of oilseed rape: taking oilseed rape at the seedling stage as the test object,the wind oscillation images of oilseed rape in the horizontal direction can be effectively obtained through the UAV optical camera in the field test.The vibration frequency of oilseed rape in the horizontal direction was not significantly different from that in the vertical direction,while the amplitude in the horizontal direction was significantly smaller than that of oilseed rape in the vertical direction.(3)Static indices such as bending mechanical properties of rape and rape vegetation index were extracted to study the relationship between such indices and wind-driven response characteristics of rape.The modulus of leaf petiole elasticity varied from 1.2 to2.6 MPa,and the difference in the modulus of leaf petiole elasticity between different cultivation plots was significant.Among the 13 vegetation indices extracted from visible light images of rape seedlings,eight vegetation indices,namely NRI,NGI,MGRVI,RBRI,NPCI,GBRI,VDVI and EXGR,showed some correlation with wind response parameters of rape,among which only three vegetation indices,namely RBRI,EXGR and VDVI,showed significant.The vegetation indices of RBRI,EXGR and VDVI showed significant correlation with wind response parameters in all trials.(4)By reducing the dimensionality of 18 oilseed rape parameters,including wind response characteristics and vegetation indices at the seedling stage,the top five components contributed more than 85% to the overall parameters in the three trials,which could reflect the information contained in the 18 parameters.Three different models for predicting the inversion index were established using the five principal components as prediction inputs and the rape inversion index as output,namely the genetic algorithm optimised BP neural network model(GA-BP),the particle swarm algorithm optimised BP neural network model(PSO-BP)and the cuckoo algorithm optimised support vector machine model,of which the prediction accuracy of the CS-SVM model was significantly better than that of the GA-BP and The mean absolute percentage error MAPE of the CSSVM model was 11.2%,10.4% and 11.0% respectively in the three trials,while the root mean square error RMSE between its predicted and true values were 0.411,0.395 and0.402 respectively.Therefore,the use of seedling oilseed rape wind response characteristics as well as bending characteristics and The results show that the use of seedling oilseed rape wind response characteristics as well as bending characteristics and vegetation index at seedling stage can be used to predict the ability of oilseed rape to fall at maturity to some extent.