摘要： 针对复杂图像的目标检测问题，提出一种基于空间和时间差别采样的彩色图像分割方法。选定目标和背景的感兴趣区域作为候选样本，对图像空间高梯度区域像素进行采样，使用极限学习机学习得到粗分割目标。模拟人眼视觉神经网络，对差别像素进行重采样，利用新增样本更新分类模型。实验结果表明，该方法可以从复杂图像场景中有效地分割目标。

关键词: 图像分割, 极限学习机, 采样-学习策略, 熵, 时间不连续性, 空间不连续性

Abstract: Aiming at target detection problem of complex image, this paper presents a color image segmentation method based on spatial and temporal difference sampling. It locates a part of object and background as candidate regions for sampling. By sampling from high gradient pixels and learning by Extreme Learning Machine(ELM), it extracts object roughly. It simulates human visual nerve network, resamples from spatial and temporal discontinuities, and updates classification model by new sample. Experimental results show that the proposed method can extract object effectively from the complex scenes.

Key words: image segmentation, Extreme Learning Machine(ELM), learning by sampling strategy, entropy, temporal discontinuity, spatial discontinuity

中图分类号: 

• TP911.73