基于机器视觉的蛾类三维姿态中前翅间夹角计算方法

发布时间：2018-02-01 11:05

本文关键词： 机器视觉三维姿态蛾类前翅间夹角特征点提取　出处：《林业科学》2017年11期 　论文类型：期刊论文

【摘要】：【目的】在农林业害虫自动识别分类过程中,目标蛾类三维姿态的准确获取可以优化识别过程,提高识别效率。通过对复杂的蛾类害虫三维姿态进行量化,准确获取虫体三维姿态的信息数据,可克服二维姿态识别的信息缺失问题,提高算法的鲁棒性,为蛾类虫体的自动识别奠定基础。【方法】以棉铃虫为例,提出一种基于机器视觉原理的蛾类虫体前翅间夹角计算方法,以确定虫体的三维姿态,即:通过角点检测原理提取蛾类虫体前翅的标记特征点,获取标记特征点的空间坐标,进而计算虫体前翅间夹角角度。【结果】此方法能够快速、便捷、准确地获取棉铃虫成虫虫体前翅间夹角,且相对误差0.10%~3.96%;该计算方法与激光测量进行偏差分析,均方根误差为1.421 6;配对T检验无显著性差异,表明本文提出的方法可行。【结论】以棉铃虫为例提出一种基于机器视觉的标记特征点虫体前翅间夹角计算方法,平均用时仅14.6 s,少于激光测量法的1 min,在计算效率上也有所提高,为多姿态蛾类害虫的自动监测、快速识别提供重要的技术手段。
[Abstract]:[objective] in the process of automatic identification and classification of agricultural and forestry pests, the accurate acquisition of 3D attitude of target moths can optimize the identification process and improve the recognition efficiency. Accurate acquisition of three-dimensional attitude information data can overcome the problem of information deficiency in two-dimensional attitude recognition, improve the robustness of the algorithm, and lay the foundation for automatic recognition of moth. [methods] take Helicoverpa armigera as an example. A method of calculating the angle between the front wings of moth based on the principle of machine vision is proposed to determine the three-dimensional attitude of the moth, that is, to extract the marker feature points of the front wing of the moth by the principle of corner detection. The spatial coordinates of the marker feature points were obtained and the angle between the front wings of the insect body was calculated. [results] this method can quickly, conveniently and accurately obtain the angle between the front wings of the adults of Helicoverpa armigera. The relative error is 0.10 and 3.96; The deviation between the method and the laser measurement is analyzed. The root mean square error is 1.421 6. There was no significant difference in paired T test, which indicated that the proposed method was feasible. [conclusion] A new method based on machine vision for calculating the angle between the front wings of Helicoverpa armigera was proposed. The average time is only 14.6 s, which is less than one minute of the laser measurement method, and the computational efficiency is also improved, which provides an important technical means for automatic monitoring and rapid identification of multi-attitude moth pests.
【作者单位】：北京林业大学林木有害生物防治北京市重点实验室;国家农业信息化工程技术研究中心农业部农业信息技术重点开放实验室北京市农业物联网工程技术研究中心;
【基金】：北京市自然科学基金项目(4132027);北京市自然科学基金青年项目(6164034) 国家自然科学基金青年科学基金项目(31301238)
【分类号】：S433.4;TP391.41
【正文快照】： 蛾类害虫种类识别是农林害虫测报与防治工作的基础。随着计算机技术的发展,基于图像的害虫自动识别技术具有省时省力、智能化等优点,有助于提高害虫识别效率(姚青等,2011)。目前广泛应用二维图像技术进行害虫的自动监测识别,但由于获取到的二维图像仅对图像中虫体的颜色、几何

【相似文献】