| The utmost importance of mastering the distribution information of the irrigation canal system is to augment the effectiveness of agricultural water transmission and distribution,dynamic monitoring of irrigation water regime,and decision-making management in large-scale irrigation areas.The UAV is chosen as the remote sensing platform for this paper,and the visible camera and multispectral camera are both equipped to capture aerial pictures of the research area,thus producing a high-resolution image of the irrigation canal system.The distribution characteristic information of canal system is extracted from the visible remote sensing image and multispectral remote sensing image of the study area respectively.The accuracy of the extraction results is evaluated through experimental verification,with two aspects of integrity and accuracy taken into account.The following are the primary research findings and conclusions of this paper:(1)Investigation of the distribution of significant irrigation canal systems in the study area from the ground was conducted using two UAVs,one with a visible camera on its fixed wing and the other with a multispectral camera on its four rotor.Aerial remote sensing images were then acquired and preprocessed,with analysis revealing the color characteristics of irrigation canal.system in the image are dark and the brightness is dark;The shape feature shows that the curvature change is not obvious linear in a certain range;After the spectral characteristics are calculated in the band,different irrigation canals have spectral digital quantization values in specific intervals;Topological characteristics: the canal system has continuity and will not be interrupted suddenly.(2)K-means clustering,based on the hue of irrigation canal system invisible light remote sensing images,is employed to set the initial number of centers at 3.Iterative operation is then employed to segment the image into three categories based on color.Based on the geometric characteristics of irrigation canal system and morphological operation,post-processing such as extraction of irrigation canal system,connection of broken canal system and refinement of canal system can be realized.The results of the experiment demonstrate that the extraction accuracy and integrity of visible remote sensing images is 0.6% and 0.55 respectively.This is beneficial for small areas with a straightforward canal system distribution,yet it is not suitable for large irrigation areas with intricate ground features and intricate canal system distribution.An object-oriented classification and extraction method,based on various levels,is employed to extract the canal system from multispectral remote sensing images.This is done by utilizing the spectral and geometric features of the canal without irrigation,setting the threshold value of the characteristic parameters through the formation of rules,and then extracting the spectral features of the utilization water of the canal.By utilizing the combination of layered extraction results and the geometric properties of the canal,the vector data and distribution information of the irrigation canal system can be extracted.Rectifying the vector data will allow for the extraction of the vector results of the irrigation canal system..The results show that the integrity of the extraction results of multispectral remote sensing images is up to 0.85,with an accuracy of 0.96.The results of canal system extraction based on object-oriented classification and extraction method at different levels are ideal.(3)The extraction of complete irrigation canal system is realized by effectively combining the difference characteristics of trunk and branch canals,agricultural gross canals,canals and anhydrous canals in remote sensing images.The study area’s multispectral remote sensing images have been used to demonstrate the efficacy and practicality of this method,which can yield high-grade irrigation canal system distribution data. |
PDF Full Text Request