摘要
选择了平滑滤波、阈值分割等算法,利用图像处理技术和神经网络技术,对大豆灰斑病进行了检测,在标准豆粒与灰斑病豆粒混合的条件下计算出病粒的百分比。同时,在暗箱条件下用照相设备采集图像,利用VC++开发平台,编写程序对图像进行去噪和分割后,通过实验设计和数据统计分析提取出豆粒的23个形态特征和颜色特征参数。采用BP神经网络对豆粒进行进一步的测评。实验取得了良好的结果,识别准确,为今后大豆其他缺陷检测打下良好的基础。
The research inspected soybean frogeye spot with virtue of algorithms by image processing technology and artificial neural network （ANN）. It could also inspect soybean for spotted soybean, and compute percentage of spotted ones to the total soybeans （It could also compute the percentage of spotted ones to the total soybeans when they are mixed）. The digital image was obtained by a digital camera in the camera obscura, and then denoise and use the iterative threshold segmentation treat to images to extract 23 parameters. BP Neural Network was used to distinguish soybeans. The experiment would come off satisfactorily. It was more exact in parameter identification, it laid a good foundation for the essence for other defect inspection.
出处
《东北农业大学学报》
CAS
CSCD
北大核心
2010年第4期119-124,共6页
Journal of Northeast Agricultural University
基金
第41批中国博士后科学基金资助项目(20070410883)
黑龙江省博士后资助项目(LBHZ_07233)
东北农业大学校博士启动基金
关键词
图像处理
图像分割
BP神经网络
image process
image segmentation
BP neural network
