自适应控制下图像分割及并行挖掘算法

针对海量图像数据中目标的分割及识别问题,提出了一种自适应控制下图像分割及并行挖掘算法.采用隶属度函数窗口宽度在图像直方图控制下自适应调整模糊阈值图像分割方法对图像进行分割,提取出感兴趣的潜在目标区域,基于共轭梯度法改进的BP神经网络算法对潜在的目标区域进行训练和识别,识别算法基于OpenMP并行处理模型开发来提高执行效率.结果表明:本文算法相对于基于偏移场的模糊C均值、灰度波动变换自适应阈值和自适应最小误差阈值具有更高的分割准确率,与传统神经网络算法的识别结果相比,平均识别率提高了8%,运行时间减少了2. 5 s.

Image segmentation and parallel mining algorithm based on adaptive control

Aiming at the segmentation and recognition problem of targets in the massive image data, an image segmentation and parallel mining algorithm based on the adaptive control was proposed. The fuzzy threshold image segmentation method was used to segment the images with the adaptive adjustment of membership function window width under the control of image histogram. The potential interested target area was extracted. The BP neural network algorithm based on the conjugate gradient method was used to train and identify the potential target area, and the execution efficiency was improved with the recognition algorithm based on the development of OpenMP parallel processing model. The results show that the algorithm has higher segmentation accuracy than the fuzzy C-means based on the offset field, the adaptive threshold of gray-scale fluctuation transformation and the adaptive minimum error threshold. In comparison with the recognition results of traditional neural network algorithm, the average recognition rate increases by 8%, and the operating time reduces by 2.5 s.