| With an apple planting area of more than 30 million mu,China is the largest apple producing country in the world.Pesticide spraying is the main pest control method in apple orchard.However,excessive pesticide application not only cause environmental pollution,but also produce adverse effects such as pesticide waste and excessive pesticide residues in fruits.Precise pesticide application is an effective means to solve the above problems.Inorder to accurately control the wind power and spray dosage according to the wind power and pesticide demand at different locations,it is necessary to detect the target volume and biomass information in real time.Aiming at the problems of complex target volume detection method,poor adaptability of leaf area detection model and lack of wind regulation model in apple orchard.This paper studied the detection technology of orchard tree canopy contour and volume distribution based on laser detection and ranging(LiDAR),and designs canopy meshing profile characterization(CMPC)detection method.The detection method of canopy leaf area of orchard tree was studied,leaf area detection model was constructed by using partial least squares regression(PLSR)algorithm and BP(back propagation)neural network algorithm.The wind force in different areas of the canopy was obtained through the three-dimensional measurement test platform,the wind loss model was cauculated by the regression algorithms.The droplet deposition distribution experiment was carried out to study the droplet deposition distribution characteristics based on the characteristics of the canopy target and the wind power.The study provides the method and theoretical support for the precise variable spray control of the wind.The main research contents and conclusions of this paper are as follows:(1)The CMPC detection method was proposed for apple orchard target canopy volume.The canopy volume CMPC detection method based on LiDAR detection technology was studied.LiDAR detection mobile platform and manual measurement platform were built.Experimental research on simulated canopy and orchard tree canopy were conducted by LiDAR and manual measurement respectively.The influence of moving speed of LiDAR and grid size of CMPC detection method were analyzed and studied.The results showed that when LiDAR is static,the detection accuracy of CMPC method for simulated canopy volume is 93.3%,and when LiDAR moving speed is 1.2m/s,the detection accuracy for orchard tree canopy volume is 96.3%.There have no detect delay of LiDAR moving speed for orchard tree canopy detection,and the contour shape of canopy is basically the same.Small grid size(0.01m×0.01m-0.05m×0.05m)is suitable for describing the structural characteristics of apple tree canopy,and the grid size greater than or equal to 0.05m×0.05m is suitable for apple tree canopy contour detection.(2)The leaf area detection model is constructed based on LiDAR point cloud data.To accurately measure the leaf area in different areas of the canopy,a three-dimensional measurement platform was built to study the tree target and orchard tree canopy respectively.Dense thick,sparse thick,dense thin and sparse thin tree targets were designed for research.The canopy leaf area data was obtained through manual measurement.The point cloud data in different areas of the canopy was obtained through LiDAR detection.PLSR algorithm and BP neural network algorithm were used to study the leaf area detection model of tree target.The two-dimensional leaf area detection model of LiDAR point cloud data and the three-dimensional leaf area detection model based on LiDAR point cloud data and canopy thickness were established respectively.The results of two-dimensional leaf area detection model show that the R~2 of the models obtained by PLSR algorithm is less than 0.5,and the R~2 of the models obtained by BP neural network algorithm are 0.75,0.434,0.709 and 0.335respectively.The results of three-dimensional leaf area detection model are as follows:R~2 of the models obtained by PLSR algorithm are 0.963,0.413,0.89 and 0.27 respectively,and by BP neural network algorithm are 0.973,0.53,0.9 and 0.429 respectively.The leaf area detection model obtained by BP neural network is higher than PLSR.(3)The tree target leaf area detection model is verified on the tree canopy data,and the apple tree canopy leaf area detection model is innovatively constructed.To accurately calculate the leaf area of apple tree,the tree target leaf area detection model for orchard leaf area calculation was verified.The two-dimensional and three-dimensional leaf area detection models of orchard tree canopy were constructed.Two dimensional BP neural network,three-dimensional PLSR and three-dimensional BP neural network leaf area detection models were used for the model test.The results show that the fitting accuracy of the above three models for orchard tree leaf area is 0.289–0.706.To obtain the better orchard leaf area detection models,BP neural network regression algorithm is used to construct the two-dimensional and three-dimensional leaf area detection model of orchard tree.The detection model R~2 is 0.887 and 0.843 respectively,and the fitting accuracy of the model to the verification set is 0.709 and 0.829 respectively.It can serve apple orchard precise variable spray technology.(4)Innovatively constructed the wind loss model in the canopy of orchard tree.Aiming at the problems of many influencing factors on wind loss,lack of wind control model and difficult to measure the wind force in orchard tree canopy.The wind loss model of orchard tree was studied.The position of canopy in space was located through three-dimensional measurement platform,and the wind force at different positions in canopy was measured.The classical regression algorithm,PLSR regression algorithm and BP neural network regression algorithm was used to construct the wind loss model under different fan speeds of 1381r/min,1502r/min and 1676r/min.The results show that the accuracy of the model obtained by BP neural network regression algorithm is the highest.The R~2 of the model is 0.783,0.679 and 0.715 respectively,which is 0.18,0.173 and 0.266 higher than the classical regression algorithm,and 0.159,0.003 and 0.145 higher than the PLSR algorithm.The formula of the model obtained by the classical regression algorithm is complex,PLSR and BP neural network algorithms can be selected to regulate the wind force of fruit tree canopy according to the requirements of pesticide application.(5)The study of characteristics of droplet deposition and distribution by air assisted sprayer in canopy.based on the study of orchard tree canopy volume detection,canopy leaf area detection and wind loss model,the characteristics of droplet deposition and distribution in air spraying canopy were studied.The effects of orchard tree canopy thickness,leaf area and wind force on the deposition and distribution of droplets in the canopy were studied.Different spray fan speeds of 1381r/min,1502r/min and 1676r/min were set to spray on the orchard tree canopy.The results show that there is a significant secondary correlation between droplet coverage and deposition,the deposition is used as the index in this paper;the volume median diameter(VMD)of droplets shows abnormalities,the wind loss rate is small and the leaf area value of the corresponding area is small.Through the significance analysis,it is concluded that the influence order of canopy thickness,leaf area and canopy inlet wind force on the sum of front and back deposition of water sensitive paper and the ratio of front and back droplet deposition is as follows:inlet wind speed,leaf area and canopy thickness.Aiming at the demand of precise pesticide application in apple orchard,the target feature detection of orchard tree and the wind loss model in canopy were studied.The research results provide a scientific basis for the wind regulation of precision pesticide application in apple orchard,and play a positive role in promoting the rapid development of precision pesticide application technology. |
PDF Full Text Request