基于Python手写拼音识别

'''
k:k值
testdata:测试数据集
traindata:训练数据集
labels:分类标签
'''

def knn(k, testdata, traindata, labels):
    '''定义算法'''
    traindatasize = traindata.shape[0] #计算训练集的长度
    dif = tile(testdata,(traindatasize,1)) - traindata #将测试集扩展至训练集的长度，再求差值
    sqrdif = dif**2 #求差值的平方
    sumsqrdif = sqrdif.sum(axis=1) #求平方和
    distance = sumsqrdif**0.5 #再开根号，即所有的距离
    sorted_distance = distance.argsort() #对距离进行排序，返回排序后的索引
    count = {} #准备一个空字典，存放投票结果
    for i in range(0,k):  
        vote = labels[sorted_distance[i]] #提取索引多对应的标签值作为字典的key
        count[vote] = count.get(vote,0)+1  #票数作为字典的value
    sorted_count = sorted(count.items(),key=lambda x:x[1],reverse=True) #对最后的投票结果进行排序
    return sorted_count[0][0] #返回得票最多的标签

def img2Model(originDataPath, modelpath):
    list = os.listdir(originDataPath)
    for child in list:
        s_list = os.listdir(originDataPath + child)
        for i in s_list:

            filepath=originDataPath + child + '/' + i
            # print(filepath)
            img = cv2.imdecode(np.fromfile(filepath, dtype=np.uint8),0)
            img = cv2.resize(img, (32, 32))
            img = np.asarray(img)
            img[img > 127] = 255
            img[img <= 127] = 1
            img[img == 255] = 0
            dstFileName = modelPath + i.split('.')[0] + '.txt'
            np.savetxt(dstFileName, img, fmt='%d', delimiter=' ')

# 随机分拣出测试集，其他文件为训练集
def shutildata(modelpath, trainpath, testpath):
    txtlist = os.listdir(modelpath)
    index = [random.randint(0, len(txtlist)) for i in range(10)]
    # print(index)
    arr = [txtlist[i].split('.')[0].split("_")[1] for i in index]
    for i in txtlist:
        try:
            if i.split(".")[0].split("_")[1] in arr:
                shutil.copy(modelpath + "/" + i, testpath)
            else:
                shutil.copy(modelpath + "/" + i, trainpath)
        except:
            pass

# 加载数据

def load_data(dataFilePath):
    arr = np.loadtxt(dataFilePath, dtype=np.int)
    arr = arr.flatten()
    return arr

# 建立训练数据集
def makeTrainData(trainpath):
    labels = []
    trainfile = os.listdir(trainpath)

    trainarr = np.zeros((len(trainfile), 1024))
    for i in range(0, len(trainfile)):
        # print(trainfile[i])
        thislabel = trainfile[i].split(".")[0].split("_")[0]

        if len(thislabel) != 0:
            labels.append(int(thislabel))
        trainarr[i, :] = load_data(trainpath + trainfile[i])
    return trainarr, labels

# 验证
def validate(testpath, trainpath, k):
    trainarr, labels = makeTrainData(trainpath)
    testfiles = os.listdir(testpath)
    count = 0

    # 读取字典表
    with open('num_char.JSON', 'r') as f:
        dict = json.loads(f.read())
        # print(dict)

    for i in range(0, len(testfiles)):
        testpicname = testfiles[i].split("_")[0]
        testarr = load_data(testpath + testfiles[i])
        result = knn(k, testarr, trainarr, labels)

        testpicname = dict[str(testpicname)]
        result = dict[str(result)]

        print("真正字母:"+testfiles[i] +"  " + testpicname + "  " + "测试结果为:{}".fORMat(result))
        if str(testpicname) == str(result):
            count += 1
    print("-----------------------------")
    print("测试集为:{}个,其中正确了{}个".format(len(testfiles),count))
    print("正确率为{}".format(count / len(testfiles)))
    print()