Research On Temperature Extraction Method Of Maize Canopy By Adaptive Canny Edge Detection Based On Surface Blur

Rapid,non-destructive and accurate access to crop canopy temperature is an important indicator to monitor whether crops are affected by adverse environmental factors such as drought,pests and diseases,and help farmers to develop more reasonable irrigation and fertilization programs for crops,so as to achieve the goal of increasing income and increasing production.The problem of time,effort and range limitation of obtaining canopy temperature by artificial hand-held canopy thermometer and satellite remote sensing is that UAV can acquire remote sensing images in a wide range and quickly,and use thermal infrared images as single data.The source is more difficult to extract high precision crop canopy temperatures.This paper takes corn as the research object,and uses the six-rotor UAV to carry the thermal imager and the Dajiang Elf4 Pro UAV to obtain remote sensing images.High-precision orthophoto,thermal infrared image and measured canopy temperature were used as data sources.The watershed algorithm,Support Vector Machines(SVM)and improved Canny edge detection method were used to segment the orthophotos to obtain corn.The canopy area,and based on this,the corn canopy temperature is extracted in a thermal infrared image.The main research contents and results are as follows:(1)Corn canopy temperature extraction method based on watershed algorithm.In this paper,the watershed algorithm is used to calculate the segmentation function,Marking foreground and background with morphological-based opening and closing reconstruction algorithm,modify the segmentation function and the watershed transform to segment the orthophoto,extract the corn canopy region and generate the surface vector file,which is masked in the thermal infrared image.Membrane to extract corn canopy temperature.The experimental results show that the average accuracy,completeness and accuracy of the corn canopy segmented by this method are 73.54%,71.41%and 54.78%,respectively.The extracted canopy temperature and measured values of RMSE and R~2 are respectively 2.1850 and 0.7231.The method has a sensitive response to weak edges in the segmentation process,and the gray scale mutation on the image surface will cause a certain over-segmentation phenomenon.,which makes the extracted canopy temperature accuracy low,and it is difficult to meet the requirements of fine research.(2)Corn canopy temperature extraction method based on support vector machine.A reliable training sample is obtained by manual selection.Obtain reliable training samples by manual selection,select Radial Basis Function(RBF),optimize the kernel parameters by using the grid parameter optimization method,and learn the SVM model..Based on this,a mask is obtained in the thermal infrared image.To extract the corn canopy temperature.The experimental results show that the average accuracy,completeness and accuracy of the corn canopy region obtained by this method are 87.09%,84.41%and 73.88%,respectively.The extracted canopy temperature and measured values are RMSE and R~2,respectively.For 1.5896 and 0.8592.The method can segment the corn canopy area,However,the ability to segment corn canopy and weeds is poor during the segmentation process,but there will be leakage segmentation and multi-segmentation,resulting in incomplete or more non-canopy regions of the extracted corn canopy.(3)Corn canopy temperature extraction method based on improved Canny edge detection.Aiming at the problem of SVM and watershed algorithm in the segmentation process of corn canopy segmentation and over-segmentation,the traditional Canny edge detection algorithm is improved,that is,the selective surface blur is used instead of Gaussian blur to solve the problem of smooth edges.The optimal gradient operator is used to solve the problem that the traditional operator is very sensitive to noise.The adaptive synchronization search high and low threshold based on the maximum inter-class variance algorithm is used to solve the problem that it is difficult to search for the optimal double threshold in the traditional Canny operator.The improved Canny edge detection algorithm was used to classify the orthophoto images to obtain the corn canopy area,and the thermal infrared image was used to extract the corn canopy temperature.The experimental results show that the average accuracy,completeness and accuracy of the corn canopy region extracted by this method are 93.88%,92.97%and 87.98%,respectively.The extracted canopy temperature and measured values are RMSE and R~2,respectively.It is 1.1982 and 0.9257.The method can better segment the corn canopy region,so that the extracted corn canopy temperature accuracy is higher.