基于随机森林的大豆外观品质识别的研究

Soybean Appearance Quality Detection and Identification Based on Random Forests

不同等级的大豆外观质量与其内部营养等级存在一定关系,因此快速、精准地识别大豆病态种类至关重要。模式识别方法众多,本文采用随机森林方法进行研究。选取相应的大豆籽粒图像对其进行处理,从中挑选1 0幅图像,提取其形态特征,颜色特征,纹理特征,应用随机森林方法建立大豆外观品质识别模型,然后对大量样本进行试验。试验结果表明:不同种类病害大豆要想达到理想结果,训练步数各不同。该方法具有鲁棒性好、准确度高及系统稳定等特点。

There are certain relationship for different levels of soy appearance quality and their internal nutrition level. So it is very important for fast and accurate detection soybean appearance quality. And there are a lot of pattern recognition method,the article adopts random forests to study. It selects corresponding grains of soybean that make image processing,choose the 10 images and extract 8 morphological characteristics variables to establish soybean appearance test model. It showed that if diseased soybean of different species need to achieve ideal result,it must make different training.That is concluded that this method is high accuracy and system stability finally.