极限学习机是由黄广斌在2005年作为一种新的单隐层前馈神经网络提出的，具有与神经网络（NN）相同的全局逼近性质，且其参数学习无需迭代，速度明显快于现有的神经网络。目前在岩性识别、LF终点温度软测量、穿孔机导盘转速测量、软测量建模、图像识别等方面有所应用，但将其用于图像分割中的应用还较少。-Extreme learning machine is by Huang guangbin in 2005 as a new single hidden layer feedforward neural network is presented. And with neural network (NN) with the same global approximation properties, and learning its parameters without iteration and significantly faster than the existing neural networks. At present in lithology identification, lf end temperature soft measurement, punch guide disc speed measurement, soft measurement modeling, image recognition and so on some aspects of applications, but will be used for image segmentation is little used.