Research On Estimation Of Regional Straw Coverage Based On Scattered Sample Images

Returning straw to the field can reduce environmental pollution and waste of resources caused by burning straw.It is an important measure to protect soil and water fertility in agriculture,and plays a big role in environmental protection and sustainable development of agriculture.Because the color of the dried straw and the land is very small,the straw returning area is large,image recognition obtained by remote sensing or aerial photography is poor,cannot be used for the statistical operation of the entire working area,and by the traditional manual measurement is time consuming and laborious.Therefore,after detecting the coverage of a single straw-covered plot image,the K-nearest neighbor algorithm based on plowing direction is used to estimate the straw coverage of Regional land.The research mainly includes straw identification and coverage calculation of dispersed images and estimation of regional straw coverage and its system implementation.In the part of recognition and coverage calculation of scattered images,the theoretical basis of SVM image classification is introduced,the image of pre-processing method of energy spectrum normalization(ESN)is proposed by using the smashed straw image collected under natural light as the scattered sample image,the SVM algorithm was used to classify the straw and soil,the parameters were explored,and the morphological image processing algorithm was used to obtain accurate image coverage,the simulation process was realized by MATLAB math toolbox.In the estimation part of regional straw coverage,a K-nearest neighbor(KNN)algorithm based on the plowing direction is proposed,the straw coverage of the obtained dispersed image and its position information are used as a land cover with known coverage and position information.,to estimate the coverage of other unknown plots around,compare and count the results,and obtain the coverage of the straw area including these scattered samples,the system is implemented by C#programming through Visual Studio 2017 environment.The experiment results show that the SVM algorithm is used to classify the image of the crushed straw by ESN-processed,which has a good recognition effect,the coverage accuracy calculated by further processing the image with morphological thought is high,moreover,and all the ESN-processed images are predicted by SVM using the training set obtained through the training process,the result is not affected by illumination,and the efficiency is high.According to the verification scheme,KNN algorithm based on plowingdirection is suitable for estimating regional straw coverage,the error of estimating results is less than 5% when the conditions of dispersive density and plowing direction are satisfied.The estimation results can be realized quickly and conveniently by using the regional straw coverage estimation research system.