基于多特征融合的水果病虫害知识视频镜头检测算法  

Video shot detection algorithm of fruit pest knowledge based on multi feature fusion

作　　者：刘立 丰洪才 黄清 LIU Li;FENG Hong-cai;HUANG Qing(School of Mathematics and Computer Science,Wuhan Polytechnic University,Wuhan 430023,China;Network and Information Center,Wuhan Polytechnic University,Wuhan 430023,China;Department of Information Technology,Wuhan Dongxihu Vocational and Technical School,Wuhan 430023,China)

机构地区：[1]武汉轻工大学数学与计算机学院,武汉430023 [2]武汉轻工大学网络与信息中心,武汉430023 [3]武汉市东西湖职业技术学校信息技术系,武汉430023

出　　处：《湖北农业科学》2022年第20期195-203,共9页Hubei Agricultural Sciences

基　　金：湖北省教育厅重点科研计划(D20101703)。

摘　　要：为了提高农业知识视频镜头检测的准确性,以水果病虫害知识视频为例,运用一种多特征融合的自适应双阈值快速镜头检测方法。该方法分为初检和复检两个阶段,初检阶段根据人类视觉特点采用改进分块提取颜色特征,并结合自适应跳略方法筛选出镜头边界候选帧序列;复检阶段利用多特征融合方法对初检得到视频镜头边界进一步细化检测。结果表明,该方法能够有效实现水果病虫害知识视频镜头的快速分割,查全率大于93.5%、查准率大于93.4%,高效准确找出水果病虫害的镜头,提高农民对农业知识视频的利用率。In order to improve the accuracy of shot detection in agricultural knowledge video,taking fruit pest knowledge video as an example,an adaptive double threshold fast shot detection method based on multi-feature fusion was used. The method was divided into two stages:initial inspection and reinspection. In the initial inspection stage,the improved block was used to extract the color features according to the characteristics of human vision,and the candidate frame sequence of the shot boundary was selected combined with the adaptive skip method;in the inspection process,the multi feature fusion method was used to further refine the video shot boundary obtained from the initial inspection. The results showed that this method could effectively realize the rapid segmentation of fruit diseases and insect pests knowledge video shot,with a recall rate of more than 93.5% and a precision rate of more than 93.4%. It could efficiently and accurately find the shot of fruit diseases and insect pests,and improve the utilization rate of farmers’ agricultural knowledge video.

关 键 词：水果病虫害知识视频 镜头分割 镜头相似性 多特征融合 双阈值 

分 类 号：S436.6[农业科学—农业昆虫与害虫防治]