文章目录
概述KNN算法原理KNN二维分类器模型DSIFT手势识别应用手势识别具体流程概述
本文介绍了KNN算法的基本原理，以及配合dfift（稠密sift）进行一个手势识别方面的应用
KNN算法原理
KNN算法（K-Nearest Neighbor,K邻近分类法）。看似十分神秘以及高大上，实则相当简单。举一个简单的例子，在一本字典里，有很多不同的单词，它们可能对应着相同的中文意思，比如安静，在字典里它可能是“quiet”，“slience”等等。视觉图像的特征也是如此，描述视觉特征的一些信息可能属于同一类，我们可以给它们打上标签，然后储存在计算机的字典里。而在python里有这么一个用法，就叫做字典。配合字典的一些功能特性，我们可以实现KNN算法。如果对于字典的用法比较陌生，这里有一篇CSDN社区其他博主写的一篇文章，可以参考一下。
/qq_40678222/article/details/83448740
KNN二维分类器模型
# -*- coding: utf-8 -*-from numpy.random import randnimport picklefrom pylab import *# create sample data of 2D pointsn = 200# two normal distributionsclass_1 = 0.6 * randn(n,2)class_2 = 1.2 * randn(n,2) + array([5,1])labels = hstack((ones(n),-ones(n)))# save with Pickle#with open('points_normal.pkl', 'w') as f:with open('points_normal.pkl', 'w') as f:pickle.dump(class_1,f)pickle.dump(class_2,f)pickle.dump(labels,f)# normal distribution and ring around itprint "save OK!"class_1 = 0.6 * randn(n,2)r = 0.8 * randn(n,1) + 5angle = 2*pi * randn(n,1)class_2 = hstack((r*cos(angle),r*sin(angle)))labels = hstack((ones(n),-ones(n)))# save with Pickle#with open('points_ring.pkl', 'w') as f:with open('points_ring.pkl', 'w') as f:pickle.dump(class_1,f)pickle.dump(class_2,f)pickle.dump(labels,f)print "save OK!"
这里进行了两次保存，第二次需要改一下代码中的文件名，这样我们可以的到一个训练集以及一个测试集。
我们可以利用一段代码来可视化KNN算法的分类效果
# -*- coding: utf-8 -*-import picklefrom pylab import *from PCV.classifiers import knnfrom PCV.tools import imtoolspklist=['points_normal.pkl','points_ring.pkl']figure()# load 2D points using Picklefor i, pklfile in enumerate(pklist):with open(pklfile, 'r') as f:class_1 = pickle.load(f)class_2 = pickle.load(f)labels = pickle.load(f)# load test data using Picklewith open(pklfile[:-4]+'_test.pkl', 'r') as f:class_1 = pickle.load(f)class_2 = pickle.load(f)labels = pickle.load(f)model = knn.KnnClassifier(labels,vstack((class_1,class_2)))# test on the first pointprint model.classify(class_1[0])#define function for plottingdef classify(x,y,model=model):return array([model.classify([xx,yy]) for (xx,yy) in zip(x,y)])# lot the classification boundarysubplot(1,2,i+1)imtools.plot_2D_boundary([-6,6,-6,6],[class_1,class_2],classify,[1,-1])titlename=pklfile[:-4]title(titlename)show()
DSIFT
dsift（dense—sift），稠密sift，其特点如字面意思，它是密集采样的sift描述子。具体操作过程可以参考下面这篇博客
/langb/article/details/48738669
这里我们利用对一张图片进行了DSIFT特征提取，并记录了它的可视化结果
# -*- coding: utf-8 -*-from PCV.localdescriptors import sift, dsiftfrom pylab import *from PIL import Imagedsift.process_image_dsift('gesture/empire.jpg','empire.dsift',90,40,True)l,d = sift.read_features_from_file('empire.dsift')im = array(Image.open('gesture/empire.jpg'))sift.plot_features(im,l,True)title('dense SIFT')show()
可视化结果：
手势识别应用
我们要利用上面提到的KNN算法以及DSIFT，实现手势识别
手势识别具体流程
提取dsift特征利用KNN算法构建“字典”使用测试集训练模型这是手势意义以及对应dense-sift的可视化# -*- coding: utf-8 -*-import osfrom PCV.localdescriptors import sift, dsiftfrom pylab import *from PIL import Imageimlist=['gesture/train/C-uniform02.ppm','gesture/train/B-uniform01.ppm','gesture/train/A-uniform01.ppm','gesture/train/Five-uniform01.ppm','gesture/train/Point-uniform01.ppm','gesture/train/V-uniform01.ppm']figure()for i, im in enumerate(imlist):print imdsift.process_image_dsift(im,im[:-3]+'dsift',10,5,True)l,d = sift.read_features_from_file(im[:-3]+'dsift')dirpath, filename=os.path.split(im)im = array(Image.open(im))#显示手势含义titletitlename=filename[:-14]subplot(2,3,i+1)sift.plot_features(im,l,True)title(titlename)show()
下面是针对我们一个训练集的训练结果
# -*- coding: utf-8 -*-from PCV.localdescriptors import dsiftimport osfrom PCV.localdescriptors import siftfrom pylab import *from PCV.classifiers import knndef get_imagelist(path):""" Returns a list of filenames forall jpg images in a directory. """return [os.path.join(path,f) for f in os.listdir(path) if f.endswith('.ppm')]def read_gesture_features_labels(path):# create list of all files ending in .dsiftfeatlist = [os.path.join(path,f) for f in os.listdir(path) if f.endswith('.dsift')]# read the featuresfeatures = []for featfile in featlist:l,d = sift.read_features_from_file(featfile)features.append(d.flatten())features = array(features)# create labelslabels = [featfile.split('/')[-1][0] for featfile in featlist]return features,array(labels)def print_confusion(res,labels,classnames):n = len(classnames)# confusion matrixclass_ind = dict([(classnames[i],i) for i in range(n)])confuse = zeros((n,n))for i in range(len(test_labels)):confuse[class_ind[res[i]],class_ind[test_labels[i]]] += 1print 'Confusion matrix for'print classnamesprint confusefilelist_train = get_imagelist('gesture/train')filelist_test = get_imagelist('gesture/test')imlist=filelist_train+filelist_test# process images at fixed size (50,50)for filename in imlist:featfile = filename[:-3]+'dsift'dsift.process_image_dsift(filename,featfile,10,5,resize=(50,50))features,labels = read_gesture_features_labels('gesture/train/')test_features,test_labels = read_gesture_features_labels('gesture/test/')classnames = unique(labels)# test kNNk = 1knn_classifier = knn.KnnClassifier(labels,features)res = array([knn_classifier.classify(test_features[i],k) for i inrange(len(test_labels))])# accuracyacc = sum(1.0*(res==test_labels)) / len(test_labels)print 'Accuracy:', accprint_confusion(res,test_labels,classnames)
修改一行代码：
dsift.process_image_dsift(filename,featfile,10,5,resize=(100,100))，把第一次中的50 * 50改为100 * 100，可以得到如下结果：
在改为100 * 100后，准确率提高了约8个百分点，在原大小下“p”和“v”的较高混淆率也得到了降低。
