基于多角度图像特征的果蔬识别方案设计-以西红柿为例

Design of fruit and vegetable recognition scheme based on multi angle image features-taking tomato as an example

中文关键词:  果蔬  多角度特征  Haar-like  AdaBoost分类器  识别

英文关键词:fruits and vegetables  angle feature  Haar-like  AdaBoost classifier  recognition

中文摘要:

      目的 设计基于多角度 像特征的果蔬识别方案。方法 采用多角度特征的果蔬识别算法, 在水平、垂直、多角度旋转等Haar-like特征的基础上与AdaBoost自学习算法充分结合。通过离线训练, 获得识别西红柿的AdaBoost分类器, 在此基础上以平均像素值为核心创造颜色特征分类器, 使Haar-like与AdaBoost分类器有机结合, 实现对果蔬类型的自动识别。结果 以西红柿为例进行识别时, 准确性超过95%, 且该方法对干扰因素具有较强的抗性, 完成一帧图像的识别只需85 ms。结论 该方法能够迅速的完成识别任务, 达到了实时性方面的要求。

英文摘要:

      Objective To design a fruit and vegetable recognition scheme based on multi-angle image features. Methods The fruit and vegetable recognition algorithm based on multi-angle features was combined with AdaBoost self-learning algorithm on the basis of horizontal, vertical and multi-angle rotation Haar-Like features. Through offline training, an AdaBoost classifier for tomato recognition was obtained. On this basis, a color feature classifier was created based on the average pixel value. The Haar-like classifier and AdaBoost classifier were combined organously to realize automatic recognition of fruit and vegetable types. Results When tomatoes were used as an example, the accuracy was more than 95%, and the method had strong resistance to interference factors, it only took 85 ms to complete the recognition of one frame of image. Conclusion This method can quickly complete the recognition task, which meets the real-time requirements.