曾窕俊,吴俊杭,马本学,汪传建,罗秀芝,王文霞.基于帧间路径搜索和E-CNN的红枣定位与缺陷检测[J].农业机械学报,2019,50(2):307-314.
ZENG Tiaojun,WU Junhang,MA Benxue,WANG Chuanjian,LUO Xiuzhi,WANG Wenxia.Localization and Defect Detection of Jujubes Based on Search of Shortest Path between Frames and Ensemble-CNN Model[J].Transactions of the Chinese Society for Agricultural Machinery,2019,50(2):307-314.
摘要点击次数: 1937
全文下载次数: 1090
基于帧间路径搜索和E-CNN的红枣定位与缺陷检测   [下载全文]
Localization and Defect Detection of Jujubes Based on Search of Shortest Path between Frames and Ensemble-CNN Model   [Download Pdf][in English]
投稿时间:2018-08-10  
DOI:10.6041/j.issn.1000-1298.2019.02.035
中文关键词:  红枣  定位  缺陷检测  路径搜索  卷积神经网络
基金项目:国家自然科学基金项目(61561041、61763043)、国家重点研发计划项目(2017YFB050420)和东南大学计算机网络和信息集成教育部重点实验室开放课题项目(K93-9-2018-10)
作者单位
曾窕俊 石河子大学 
吴俊杭 石河子大学 
马本学 石河子大学 
汪传建 石河子大学 
罗秀芝 石河子大学 
王文霞 石河子大学 
中文摘要:针对红枣自动分级视频图像中红枣定位、缺陷检测难问题,提出一种基于帧间最短路径搜索的目标定位方法和集成卷积神经网络模型(Ensemble-convolution neural network,E-CNN)。通过建立图像坐标系及图像预处理,获得图像中单个红枣目标的位置坐标,并将其映射至空间坐标系中,结合帧间最短路径判定规则,实现目标位置坐标随视频时间序列更新、传递,并且运用此方法快速、有效地构建数据集。基于“Bagging”集成学习方式,采用E-CNN通过训练集构建基础卷积神经网络树模型,再根据每棵基础树模型输出结果,通过“投票”方式得出模型最终结果。试验结果表明,利用帧间最短路径搜索的目标定位方法,定位准确率达100%。同时,使用E-CNN,模型的识别正确率和召回率分别达到98.48%和98.39%,分类精度大于颜色特征分类模型(86.62%)、纹理特征分类模型(86.40%)和基础卷积神经网络模型(95.82%)。E-CNN模型具有较高的识别正确率及较强的鲁棒性,可为其他农产品分选、检测提供参考。
ZENG Tiaojun  WU Junhang  MA Benxue  WANG Chuanjian  LUO Xiuzhi  WANG Wenxia
Shihezi University,Shihezi University,Shihezi University,Shihezi University,Shihezi University and Shihezi University
Key Words:jujube  location  defect detection  path search  convolution neural network
Abstract:Jujube is a high-value fruit throughout the world. Detection for the surface defects of jujubes is the prerequisite to realize its automatic grading. For the difficulty of red jujube location and defect detection in images, a kind of locating method based on search of the shortest path between frames and framework named ensemble-convolution neural network (E-CNN) were introduced. As for locating method, image coordinate system was established at first. With image preprocessing, the location coordinates of each jujube target were obtained and these location coordinates were mapped into the spatial coordinate system. Combining judgment rules of the shortest path between frames, the location coordinates of targets were updated and transmitted with the video time sequences. Also, by using the method with video data, it could be quickly and efficiently to build data sets. Based on “Bagging” ensemble learning and “returning” training method, the basic convolutional neural network tree models were built and then according to output of each basic tree model, the final result of the model was obtained by “voting”. The results of experiments showed that the location accuracy of 100% was achieved with this locating method, avoiding complicated mechanical and circuit design. At the same time, by using E-CNN model, the average recognition accuracy and recall rate reached 98.48% and 98.39%, respectively. And the classification accuracy was greater than those of color feature classification model (86.62%), texture feature classification model (86.40%), and basic convolution neural network model (95.82%). The model had high recognition accuracy and strong robustness, and can provide reference for other agricultural products sorting and detection.

Transactions of the Chinese Society for Agriculture Machinery (CSAM), in charged of China Association for Science and Technology (CAST), sponsored by CSAM and Chinese Academy of Agricultural Mechanization Science(CAAMS), started publication in 1957. It is the earliest interdisciplinary journal in Chinese which combines agricultural and engineering. It always closely grasps the development direction of agriculture engineering disciplines and the published papers represent the highest academic level of agriculture engineering in China. Currently, nearly 8,000 papers have been already published. There are around 3,000 papers contributed to the journal each year, but only around 600 of them will be accepted. Transactions of CSAM focuses on a wide range of agricultural machinery, irrigation, electronics, robotics, agro-products engineering, biological energy, agricultural structures and environment and more. Subjects in Transactions of the CSAM have been embodied by many internationally well-known index systems, such as: EI Compendex, CA, CSA, etc.

   下载PDF阅读器