基于视觉识别技术的稻粒特征识别检测系统  

Rice Grain Feature Recongnition and Detection System Based on Visual Recognition Technology

作　　者：刘春山[1] 李金琼 陈思羽[1] 刘洪义[2] 潘佳琦 焦仁宝[1] 初旭宏[1] Liu Chunshan;Li Jinqiong;Chen Siyu;Liu Hongyi;Pan Jiaqi;Jiao Renbao;Chu Xuhong(School of Mechanical Engineering,Jiamusi University,Jiamusi 154007,China;Heilongjiang Academy of Agricultural Machinery Sciences,Harbin 150081,China)

机构地区：[1]佳木斯大学机械工程学院,黑龙江佳木斯154007 [2]黑龙江省农业机械工程科学研究院,哈尔滨150081

出　　处：《农机化研究》2025年第5期138-144,共7页Journal of Agricultural Mechanization Research

基　　金：黑龙江省自然科学基金联合引导项目(LH2023C059)。

摘　　要：通过视觉识别图像处理技术对稻谷谷粒进行处理,利用OTSU自适应阈值对稻谷图像进行二值化处理,采用中值滤波对图像进行降噪处理、图像分割和连通区域标记,基于几何形状特征阈值检测算法识别不同种类稻谷的长度、宽度、周长和面积像素值,提出一种综合的阈值通用旋转法,可实现不同谷粒的分类。根据获取谷粒图像样品和采集谷粒数目特征,基于形态学处理和分水岭变换分割算法对谷粒进行数目识别提取,最后根据检测的数目和整碎谷粒特征统计汇总稻谷谷粒的相关参数信息,将谷粒识别结果和相关数据信息保存在数据库中,便于进一步分析。研究结果表明:利用MatLab软件中GUI界面开发的稻谷检测系统对谷粒数目有较好的识别力,精确度达98.327%,对于整碎谷粒的识别误差在可控范围,满足要求与传统人工检测相比,具有检测快速、多样化等优点。Processing rice grain by visual recognition image processing technology,binarizing the rice grain image using OTSU adaptive thresholding,using median filtering for image noise reduction processing,image segmentation and connected region labeling,identifying the length,width,perimeter and area pixel values of different kinds of rice grain based on geometric shape feature threshold detection algorithm,and proposimg a comprehensive threshold to achieve the classification of different grains.Based on obtaining grain image samples and collecting grain number features,the number identification extraction of grains was performed based on morphological processing and watershed transform segmentation algorithm,and finally the information of relevant parameters of rice grains was summarized according to the detected number and whole broken grain feature count,and the grain identification results and related data information were saved in the database for further analysis.The results showed that the rice grain detection system developed by using the GUI interface in MatLab software had good recognition power for the number of grains,its accuracy reached 98.327%,and the recognition error for whole broken grains was within the controllable range,which met the requirements compared with the traditional manual detection,had the advantages of rapid detection and diversity.

关 键 词：检测系统 特征识别 图像处理 稻粒 

分 类 号：S126[农业科学—农业基础科学]