手势图像的特征提取与识别

手势图像的特征提取与识别(任务书,开题报告,外文翻译,论文11000字)
摘要
     对于日益增长的人机交互市场，本文对一定环境条件下，一个固定摄像头的手势识别情形进行了研究，利用手势分割建立手势库，手势的特征提取和识别。噪声所产生的因素，如采集装置、环境影响等因素，采用中值滤波法进行滤波。滤除噪声后，转化成灰度图像，并基于肤色和背景减法的自适应阈值分割法分割图像。图像分割边缘检测和手势形状特征提取的图像特征提取。Log边缘检测算法提取图像边缘分割的手势。[1]针对不变矩的旋转、平移和缩放不变性，给出了利用矩特征提取手势轮廓的方法。模板库是选取网络上无噪声污染的标准手势图像，将在理想的条件下采集到的图片的手势定位，LOG边缘检测算法提取边缘特征，然后填充，定位以及分割。之后做成匹配模板。在自然环境的提取后，利用该方法计算了模板特征，并利用最近邻原则作为识别准则。最后，中值滤波、空间变换、Log边缘检测、形状匹配进行了验证，手势识别和跟踪的结果进行了分析。得出以下结论：对自然场景的采集手势可以完全分离，并去除大部分噪声的图像；手跟踪稳定效果，可以找回丢失的跟踪。
关键词：定位分割；形状匹配；Log边缘检测；

Abstrct
      For the growing market for human-computer interaction, this paper under the certain conditions, a stationary camera gesture recognition situation were studied use of gesture segmentation to establish a gesture library, gesture feature extraction and recognition. Noise generated by factors, such as the acquisition device, environmental impact and other factors, using the median filtering method to filter. After filtering out the noise, the image is transformed into gray level image, and the image is segmented by adaptive threshold segmentation method based on skin color and background subtraction. Image segmentation edge detection and gesture shape feature extraction of image feature extraction. Log edge detection algorithm to extract image edge segmentation gesture. According to the invariance of rotation, translation and scale invariance of invariant moments, the method of extracting gesture contour using moment feature is presented. Template library is selected on the network without noise pollution standard gesture image, will be in ideal conditions collected pictures hand localization and log edge detection algorithm for extracting edge features, then filled with positioning and segmentation. Then make a matching template. After the extraction of the natural environment, this method is used to calculate the template feature, and the nearest neighbor principle is used as the recognition criterion. Finally, median filtering, spatial transform, Log edge detection, shape matching are verified, and the results of hand gesture recognition and tracking are analyzed. Draw the following conclusion: the natural scene of the acquisition of hand gestures can be completely separated, and the removal of most of the noise image; hand tracking stability effect, you can retrieve the lost track.
Key words: location segmentation; shape matching; Log edge detection
 

目 录
第1章 绪论    4
1.1研究目的    4
1.2国内外研究现状    4
1.3研究的难点    5
1.4手势的定义    5
1.4.1手势的分类    5
1.4.2 手势系统的组成    6
1.4.3 与其他技术优势    6
1.5小结    7
第2章 手势数据库的建立    8
2.1 手势图像分割    8
2.2灰度变换    8
2.3边缘检测    9
2.3.1差分边缘检测方法    10

2.3.2 Roberts边缘检测算子    10
2.3.3 Soble边缘检测算子    10
2.3.4 Prewitt边缘检测算子    11
2.3.5 Canny边缘检测算子    11
2.3.6 Log边缘检测算子    11
2.4 手势图像形态学处理    12
2.5 本章小结    16
第3章 手势的对比    17
3.1 手势图像的平滑处理    17
3.2 手势匹配方法    18
第4章 总结    22
参考文献    23
附录    24
致谢    33