| In the growth cycle of maize,accurate and rapid acquisition of maize plant height and number can better grasp the growth information of maize,which plays a crucial role in the development of precision agriculture.Unmanned aerial vehicles(uavs)can provide important technical means for rapid,accurate and dynamic monitoring of a wide range of agricultural information,and effectively make up for some defects of ground survey.Therefore,the use of UAV remote sensing technology to quickly,accurately and widely monitor the information of maize plant height and number is of great significance for its growth monitoring and yield prediction.The different digital orthophoto map(DOM)and digital surface model(DSM)of maize with large area and high precision obtained by UAV were used to extract plant height and number,and the comparative analysis and precision evaluation were carried out with measured data.Finally,a relatively complete extraction scheme for plant height and number of maize was established.The main research contents and results are as follows:(1)Maize plant height extraction method.K-means algorithm,genetic neural network algorithm and skeleton algorithm were used to extract the maize plant area in DOM image respectively,and the plant height information was extracted by superimposing mask and DSM data.Compared with the measured plant height,the R~2 of the three methods was 0.853,0.877,0.923,RMSE was 15.886cm,14.519cm,11.493cm,MAE was 13.743cm,11.884cm8.927cm respectively.The results show that:combining DOM and DSM can better extract the height of maize in the nutritional growth stage.Compared with K-means and genetic neural network,the maize plant height extraction method based on skeleton algorithm has higher precision(R~2=0.923,RMSE=11.493cm,MAE=8.927cm).(2)Maize plant number extraction method.The threshold method and Harris corner detection algorithm were used to extract the number of maize plants in DOM images respectively.K-means and genetic neural network were used to extract the maize plants area,and threshold method was combined to extract the plant number information.Skeleton algorithm was used to extract the maize skeleton,and Harris corner detection algorithm was combined to extract the plant number information.Compared with the number of visual interpretation plants,the accuracy of the three methods was 95.94%,96.24%and 97.49%respectively.The different number between the visual interpretation and the extracted results was 42,39 and 26 respectively.The number of the same region was 12,9 and 24 respectively.The results show that:The specific number of maize experiment area can be extracted from the DOM image of UAV.Compared with the threshold method,the Harris corner detection algorithm has higher precision and more stable effect in extracting the number of maize plants,and has a good adaptability for extracting the number of maize plants under the condition of multiple plant adhesion.After extracting the skeleton of maize plants by the skeleton algorithm,the number of plants can be extracted by combining Harris corner detection algorithm with gaussian filter,which avoids the operation of removing burr and bifurcation on the skeleton for many times,saves time and has a stable effect.(3)Based on the research of(1)and(2),a high accuracy,simple and stable plant height and number extraction scheme was obtained.Skeleton algorithm was used to extract the maize skeleton in DOM image,and the height information of maize plant could be extracted by making the overlay of mask and DSM image.The extracted maize skeleton combined with Harris corner detection algorithm with gaussian filter can extract the number of maize plants.Experimental verification shows:The scheme has a high accuracy and stable extraction effect,and can provide a rapid and effective means for the extraction of plant height and number in the process of maize breeding in a wide range,and it greatly liberating manpower and promoting the development of precision agriculture. |
PDF Full Text Request