| The increasing efficiency and quantity reduction of pesticide advocated by modern agriculture provides a technical basis for the production of high quality crops.However,the spraying droplets of ultra-low dosage of pesticide on the leaves are easily restricted by the evaporation time of pesticide droplets,which seriously restricts the efficient utilization of pesticides.The evaporation research of pesticide droplets is one of the effective ways to solve this problem.In this paper,it is of great significance to study the law of droplet evaporation from three directions of image,hyperspectral and hyperspectral image fusion,in view of the key technology of droplet detection and evaporation analysis.The main research contents and conclusions are divided into 5 parts.(1)A method based on computer vision to obtain the shape boundary features of the droplet image is proposed,and the coordinates of the boundary points are obtained based on Matlab.The results of the extraction of glyphosate droplet images from the leaf of barnyard grass show that the method has a good precision for the microscopic image.An evaluation method for two-dimensional functions describing droplets is proposed based on the droplet description with two dimensional functions and the comparison of each modeling parameters.A method for dimension reduction of two-dimensional functions is proposed,and the key characteristics of droplets are analyzed using data after dimension reduction.All 11 characteristic parameters were extracted from the glyphosate droplets on the corn leaf.The results show that the method can obtain more information about the droplet detection and evaporation process through multi parameter fusion.In order to describe the change of droplet characteristics accurately,a modeling accuracy evaluation method based on statistical moments for 11 shape feature parameters is proposed.Based on the above methods,droplet detection and evaporation analysis methods based on microscopic high-speed images are proposed.The droplet detection based on the characteristic curve of the area and chord length of the image is studied.The characteristic curve of the area and the chord length can accurately obtain the characteristic peak value of the droplet,and the subtle difference of the diffusion velocity vector of the droplet in a specific direction is obtained.The results show that the characteristic value of the chord length emphasizes the description of a single point on the contour relative to the area and can be used for droplet classification.In the modeling of droplet evaporation prediction,the subtle changes in the contour of the droplet can be intuitively obtained by obtaining the coordinates of the contour points of the droplets and dividing the coordinates according to the quadrants.Based on the four parameters of eccentricity,circle rate,mean bending energy and minimum inertia axis length,a supervised linear model for the variation of droplet evaporation was established.The results showed that the model determination coefficient was between 0.7409 and 0.999 and constructed evaporation model had good accuracy.According to the results of statistical moments,it can be seen that the predicted value fluctuations,the velocity of the evaporative droplet profile change,the positive skewness and the deviation degree are all good.Statistical moment results also indicate that the modelling has reached expectations.(2)Developed a multispectral detection system for herbicide drift droplets to achieve droplet size detection and evaporation monitoring.A method based on multispectral rapid detection of droplet size of herbicides was proposed.Four bands from six bands were selected for the detection of droplets through experimental studies.The signals at-620nm and 670-690bm are the strongest.Near-infrared shortwave with wavelengths of 800-820nm and 850-870nm can be used for drop detection.In the laboratory experiment of droplet evaporation,it was found that 600-620nm wavelength was the best for the droplet evaporation detection and the correlation coefficient was 0.9353.In the field evaporation test,the multivariate linear regression model for determining droplet evaporation particle diameters at wavelengths of 600-620 nm and 670-690 nm has a model determination coefficient of 0.94.The model can well predict droplet evaporation particle size.(3)A hyperspectral imaging data method is proposed to capture droplet evaporation on time series,and the hyperspectral imaging data of three different leaf samples and foliar droplets are obtained.In the compression of spectral dimension,the SPCA is selected as the threshold condition to select the sensitive band as the threshold condition.Through the extraction and analysis of the spectral,shape and texture features of the droplets,the optimal threshold was established to construct the decision tree of droplet identification and evaporation analysis.The recognition model based on the recognition rules is built.The results of the evaporation recognition model based on spectrum,shape and texture show that the wax leaves are stable and have higher recognition precision compared with the other two kinds of blades.The shape of chorionic leaves is quite different from that of the other two kinds of leaves,but because of the larger shape of the droplets on the surface,the possibility of being mistaken is also possible.The evaporation process based on texture recognition has higher accuracy,which proves that evaporation recognition based on texture is feasible.It can effectively identify the change of droplets and droplet evaporation on different foliage,and the overall accuracy is 95.4%.The Kappa coefficient also reached 0.9472.(4)Set up an artificial simulation evaporation chamber and carry out analysis of the influence of environmental parameters(wind,temperature and humidity)on evaporation.The results showed that at the temperatures of 15 0 C,20 ℃ and 25 ℃,the turbulent wind caused evaporation time to apparently decrease by 29.15%,23.25%and 37.87%,respectively,compared to the windless conditions.The turbulent wind has the greatest impact on droplet evaporation,while the vertical wind takes the second place,and the horizontal wind has the smallest impact.The effect of wind speed on evaporation shows that the wind speed is linearly related to the evaporation time.Through the experimental study of the ambient temperature on the evaporation time,it was found that when the test temperature was set to 15℃,20℃,25℃,30℃,35℃,40℃,45℃ and 50℃ respectively,when the test glyphosate droplets with the same particle size and the ambient wind,the evaporation time of the droplets on the surfaces decreased with increasing temperature.There was a significant linear correlation between temperature and evaporation time,and the coefficient of determination of the linear regression of evaporation time for waxy leaf,flat leaf,and hairy leaf were 0.9645,0.9558,and 0.9430,respectively.Ambient humidity will increase the evaporation time of the droplets on the leaf surface,but the increase in humidity will have no significant effect on the evaporation time.The influence of the additive parameters(type and mass percentage)on evaporation was studied.The results showed that the effect of additive types on evaporation was not significant under simulated conditions.The additive mass fraction has a significant effect on evaporation.The thistle has burr leaves,and the increase of mass fraction of the addictive is very significant for the reduction of the evaporation time.However,compared with the significance of thistle,the increase of the mass fraction of additive has slightly less effect on the evaporation time of rapanea neriifolia waxy leaf surface and dandelion flat leaf surface.(5)According to the simulation of crop canopy microclimate parameters(wind speed,temperature and humidity),the instrument for measuring the evaporation time of pesticide on the leaf surface was developed,and the system accuracy was calibrated.The device can control the environmental parameters of the enclosed evaporation chamber according to the wind speed,temperature and humidity parameters set by the user,so as to provide a basis for the research of pesticide evaporation.The system wind speed control test results show that there is a good linear relationship between the duty cycle and the wind speed within the adjustable range of the PWM fan,and the determination coefficient reaches 0.9918.The PID temperature and humidity control linear model has a coefficient of determination close to 1.The results show that the instrument has good accuracy in adjusting the environmental parameters of pesticide evaporation.A direct injection variable control method was proposed for the additive.The linear correlation coefficient of the direct injection time and the direct injection dose approached 1,which indicated that the direct injection time had good accuracy for dose control.There is a good consistency between the predicted addictive quantity and the actual addictive quantity set by the controller.The linear correlation coefficient is 0.9991,which confirms that the overall accuracy of the system is very good. |
PDF Full Text Request