ass日本风韵熟妇pics男人扒开女人屁屁桶到爽|扒开胸露出奶头亲吻视频|邻居少妇的诱惑|人人妻在线播放|日日摸夜夜摸狠狠摸婷婷|制服 丝袜 人妻|激情熟妇中文字幕|看黄色欧美特一级|日本av人妻系列|高潮对白av,丰满岳妇乱熟妇之荡,日本丰满熟妇乱又伦,日韩欧美一区二区三区在线

Monitoring of planting distribution of plastic mulching farmland plays an important role in assessing the regional climate and ecological environment changes caused by it. Based on DeepLabv3+, an improved deep semantic segmentation model for plastic mulching farmland was proposed by learning the channel attention and spatial attention features for plastic mulching semantic segmentation. It can effectively segment plastic mulching farmland for unmanned aerial vehicle (UAV) multispectral remote sensing image. The UAV multispectral remote sensing images of four experimental plots during 2018—2019 were taken as the research data. The research area was Shahaoqu Irrigation Farmland in the Hetao Irrigation District, Inner Mongolia Autonomous Region. And compared with the recognition result of visible remote sensing image, by considering the appearances changes of the plastic mulching farmland, two groups of experimental schemes were designed to verify the model’s generalization performance and enhance its classification accuracy respectively. The recognition effect of the improved DeepLabv3+ semantic segmentation model was 7.1 percentage points higher than that of visible light. At the same time, the deep semantic segmentation model considering the apparent changes of mulching fields had a higher classification accuracy, and its average pixel accuracy was 7.7 percentage points higher than that without considering the apparent changes of mulching fields, indicating that the diversity of training data was helpful to improve the recognition accuracy of mulching fields. Secondly, the improved DeepLabv3+ semantic segmentation model had adaptive learning of mulch attention, in both experiments, and its classification accuracy was higher than that of the original DeepLabv3+ model. It was suggested that the attention mechanism can increase the adaptability of deep semantic segmentation model and thus improve the classification accuracy. The proposed method can accurately identify plastic mulching farmland from complex scenes and provide a method reference for monitoring plastic mulching farmland and analyzing their distribution.