基于Gentle Adaboost算法的牛体检测

目前将模式识别、机器视觉、图像处理等先进的计算机技术应用于农业信息化领域已成为当前精准养殖业的研究热点。由此提出了一种基于Gentle Adaboost算法的牛体检测的方法。首先利用Bag of Features (BOF)的思想创建特征词典,然后通过词典对目标进行特征提取,最后通过Gentle Adaboost算法对对训练集的BOF特征向量进行训练分类,获得对象和场景的分类模型。实验表明训练的检测器可精确实时的进行牛体检测。

Gentle Adaboost Based Cow Detection

At present, researches focus on the precision livestock farming which is combined with the advanced computer technology, such as pattern recognition, machine vision, image processing for the field of agricultural information. This paper presents a method for the detection of cow based on Gentle Adaboost algorithm. Firstly, feature dictionary is created by the improved BOF. Then the cow target is extracted by the dictionary. Finally, by training and classifying for the feature vector of the training set, the classification model of object and scene can be obtained. Experiments show that the detector can detect cow in real-time accurately.