摘要
近年来农作物病虫害情况呈加重态势,对粮食生产已构成直接威胁,提出了基于图像识别算法对大田作物进行病虫害数字化诊断与预警系统应用的方案设计,采用YOLOv3算法实现目标检测,引入卷积注意力模块(CBAM),通过无人机对大田作物的监测,以期提供可借鉴的方法。传统人工识别在大田信息检测上占很大劣势,其存在识别准确性低、效率低等严重缺陷,而若将图像识别算法与无人机相结合,用于检测农田,就会在降低病虫害发生概率的同时大大提高效率。经过多次实验研究,通过对无人机拍摄的大量图片进行检测对比,平均检测精度m AP从以往的60.3%上升到88.6%。监控系统利用CNN提取并融合深度光谱和局部空间特征,在大规模范围内进行精准监测。
This paper proposes a scheme design based on image recognition algorithm on how to carry out digital diagnosis and early warning system for crop diseases and insect pests in field.YOLOv3 algorithm is used to achieve target detection,and convolutional attention module(CBAM)is introduced to monitor field crops by UAV,in order to provide a reference method.Traditional manual identification has a great disadvantage in field information detection,which has serious defects such as low identification accuracy and low efficiency.However,if image recognition algorithm is combined with UAV to detect farmland,the occurrence probability of diseases and insect pests will be reduced and the efficiency will be greatly improved.After many experimental studies,the average detection accuracy mAP increased from 60.3%to 88.6%by comparing a large number of images taken by UAV.The monitoring system uses CNN to extract and fuse the depth spectrum and local spatial features for accurate monitoring in a large scale.
出处
《工业控制计算机》
2023年第4期99-100,103,共3页
Industrial Control Computer
基金
安徽省大学生创新创业省级项目基金资助(项目号S202111305117)。
