一种基于机器视觉的农田害虫活体动态图像特征获取系统  

A Dynamic Image Feature Acquisition System Based on Machine Vision

作　　者：宋琨珷 徐峰 

机构地区：[1]合肥市第三十二中学,安徽合肥230001 [2]合肥一六八中学,安徽合肥230001

出　　处：《电子质量》2017年第12期72-76,共5页Electronics Quality

摘　　要：农田害虫监测技术是实现适时、适量施药的重要前提,该技术为增加农业生产提供了重要保证。针对现有农田监测手段中运用较多的人工统计死体害虫造成的高成本、低准确性问题,该文提出一种低成本、高准确性的农田害虫活体动态特征获取系统。首先,使用间隔灯光采样法获取害虫原始图像并通过图像颜色特征和阈值法分割害虫与背景图像。然后,通过高斯滤波获得害虫的形状特征和数量信息。最后,通过帧间差分法获得害虫运动量信息。在不接触害虫的情况下完成虫情监测,减少传统监测中因害虫死体破损造成的统计误差,提高系统准确性。该系统可以为植保人员提供害虫运动的定量信息并预防大规模农作物病虫害的出现。Monitoring of farmland pests is an important prerequisite to sprinkle agriculture in a timely and app-ropriate manner,which provide an important guarantee for agricultural production.In order to solve the problem of high cost and low accuracy caused by the deadly statistics of pests,this paper presents a method for obta- ining the dynamic characteristics of pests in farmland based on machine vision.First,the original image of the pest is obtained by interval light sampling method,and the pest and the background image are segmented by image color feature and threshold method.Then,the shape and quantity information of the pests were obtained by Gaussian filtering.Finally,the pest movement information is obtained by interframe difference method.Complete monitoring insect pests without touching,Reduce the statistical error,improve system accuracy.This system can also be applied in large-scale crop pest monitoring.

关 键 词：目标轮廓检测 运动量信息获取 活体害虫识别 

分 类 号：S433[农业科学—农业昆虫与害虫防治] TP391.41[农业科学—植物保护]