• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>Pixels, Objects, Intelligence: GEOgraphic Object Based Image Analysis for the 21st Century >QUANTUM-INSPIRED EVOLUTIONARY ALGORITHM AND DIFFERENTIAL EVOLUTION USED IN THE ADAPTATION OF SEGMENTATION PARAMETERS
【24h】

QUANTUM-INSPIRED EVOLUTIONARY ALGORITHM AND DIFFERENTIAL EVOLUTION USED IN THE ADAPTATION OF SEGMENTATION PARAMETERS

机译：分段参数适应时量子启动的渐象的进化算法和差分演进

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

The key step in object-based image interpretation is segmentation. Frequently the relationship between the segmentation parameter values and the corresponding segmentation outcome is not obvious, and the definition of suitable parameter values is usually a time consuming, trial and error process. In (Costa et al., 2008), a supervised, semi-automatic method for the adaptation of segmentation parameters was proposed. Initially a human operator delineates polygons enclosing a representative set of target image objects. The manually drawn polygons are then used as reference segments by a Genetic Algorithm (GA), which searches the segmentation parameter space for values that produce segments as similar as possible to the reference. Although GA based methods have been successfully applied in many optimization problems, they are characterized by a high computational load, and do not guarantee that optimal values are found. Alternatives to the basic GA model have been proposed in order to accelerate convergence, preventing at the same time convergence to local maxima. In this work two of such alternatives have been investigated: Quantum-Inspired Evolutionary Algorithm (QIEA) (Abs da Cruz, 2007) and Differential Evolution (DE) (Storn and Price, 1997). Those two models were employed in the method proposed in (Costa et al., 2008), substituting the conventional GA originally used. Experiments showed that both models converge significantly faster than the original GA. Additionally, for an equivalent computational load, dissimilarity among the reference segments and the ones generated with the parameter values found by applying QIEA and DE was in average respectively 44percent and 50percent lower, when compared to the results obtained with the original GA.
机译：基于对象的图像解释中的关键步骤是分段。频繁，分段参数值与相应分割结果之间的关系不明显，并且合适的参数值的定义通常是耗时，试验和错误过程。在（Costa等，2008）中，提出了一种用于调整分割参数的监督，半自动方法。最初，人类运营商描绘了包含一组代表目标图像对象的多边形。然后通过遗传算法（GA）用作参考段的手动绘制的多边形，其搜索分段参数空间，以产生尽可能相似的段的值。尽管基于GA的方法已成功应用于许多优化问题，但它们的特征在于高计算负载，并且不保证找到最佳值。已经提出了基本GA模型的替代方案以加速收敛，预防同时收敛到局部最大值。在这项工作中，已经研究了两种这样的替代品：量子启发的进化算法（QIEA）（ABS Da Cruz，2007）和差分演进（DE）（粗鲁和价格，1997）。这两种模型用于（Costa等，2008）中提出的方法，替代最初使用的常规GA。实验表明，两种型号都会比原始GA更快地收敛。另外，对于等效计算负荷，参考段之间的异化性和通过应用Qiea和de发现的参数值产生的相似性平均分别为44平方，并且与用原始GA获得的结果相比，下降。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号