基于adaboost算法的模式识别.docVIP

基于adaboost算法的模式识别 摘要 模式识别技术是一种很常见的技术，它应用于各个领域，而adaboost算法是一种很经典的算法，应用也相当广泛。本文主要针对adaboost算法进行研究，这是一种迭代算法，只要对弱分类器进行研究，使弱分类器的准确度大于50%，那么由若干个弱分类器组成的强分类器便会无限接近于100%。所以判别一个弱分类器的好坏，便可以从它的错误率进行判断。 本文通MATLAB的仿真，通过对仿真结果的观察和对比，判断错误率是否符合要求，研究adaboost算法的识别能力，可以看出其研究价值，对于整个分类的过程，最重要的是需要研究其错误率的高低，如果错误率太高，那么可以认为识别识别（？），生成的强分类器的好坏在于我们对于弱分类器的选取，在选择弱分类器其参数时可在本文中看到需要注意的问题。 从辨识的结果中可以发现，使用adaboost算法生成的强分类器错误率很低，大大的提高了模式识别技术的识别能力，所以该算法能够应用于各个领域。 关键词： adaboost算法 模式识别 错误率 Abstract Pattern recognition technology is one of the most common techniques, it is applied in every field, and adaboost algorithm is a very classic algorithm, applications are very wide. This paper mainly studied adaboost algorithm for, this is a kind of iterative algorithm, but the weak classifier for study, the weak classifier, so the accuracy of more than 50% by several weak classifier composition of strong classifier will infinite close to 100%. So tell if a weak classifier is good or bad, can judgement from its error rate. This paper, through the simulation of MATLAB simulation results observation and comparison, judge whether the error rate adaboost algorithm accords with a requirement, research the recognition ability, we can see the research value, for the whole of classification procedures, the most important is to need to study its error rates the height, if the error rate is too high, so can think recognition recognition, generated strong classifier for the stand or fall of weak classifier is our selection of weak classifier, on the choice when its parameters can be in this paper to see problems needing attention. The results from identification can be found in the adaboost algorithm produces, use strong classifier error rate is very low, and greatly improve the pattern recognition technique ability to identify so this algorithm can applied in various fields Keywords： Adaboost algorithm pattern recognition error rates1 引言 1.1课题背景 长期以来，我们的生活中时时刻刻都在进行模式识别。比如我们可以认出周围的事物是树，还是花，或者是草，可以分清哪个是张三，哪个是李四。而对于这些普通