封装和使用腾讯AI开放平台的滤镜和通用OCR功能（Python）

上一篇文章封装和使用了文本情感分析功能，这一篇更进一步，封装了滤镜和通用OCR功能。
与上一篇不同的是，传递文本信息只需要将文本编码到URL中，用GET方式请求数据。图片体积图片需要先对图片进行base64编码，然后例行生成md5摘要字符串，之后使用POST方法请求数据。考虑到两种请求方式的不同，在上一篇的基础上，我对基类的接口进行了扩展，分别对两种不同的请求数据方式，设计了两个不同的接口get_url和get_params。

UML类图：

qqaibase.py

'''
date:2019.7.7
copyright:buaalzm
'''
import time
import random
import string
from abc import ABCMeta, abstractmethod


class QQAIBase():
    __metaclass__ = ABCMeta
    param_dict = {}

    def __init__(self, app_id, app_key):
        self.param_dict['app_id'] = app_id
        self.param_dict['app_key'] = app_key

    def basic_param_init(self):
        '''请求时间戳（秒级），用于防止请求重放（保证签名5分钟有效）'''
        t = time.time()
        self.param_dict['time_stamp'] = str(int(t))

        '''请求随机字符串，用于保证签名不可预测'''
        self.param_dict['nonce_str'] = ''.join(random.sample(string.ascii_letters + string.digits, 10))

    @abstractmethod
    def get_url(self):
        pass

    @abstractmethod
    def get_content(self):
        pass

    @abstractmethod
    def get_params(self):
        pass

vision_imgfilter.py

'''
date:2019.7.7
copyright:buaalzm
'''
from qqaibase import QQAIBase
import hashlib
import requests
from bs4 import BeautifulSoup
import json
from urllib.parse import quote
import base64
import random


class VisionImgFilter(QQAIBase):
    interface_url = "https://api.ai.qq.com/fcgi-bin/vision/vision_imgfilter"  # API地址
    def get_params(self, image, filter):
        self.basic_param_init()
        '''图片用base64编码'''
        base64_data = base64.b64encode(image)
        params = {'app_id': self.param_dict['app_id'],
                  'image': base64_data,
                  'time_stamp': self.param_dict['time_stamp'],
                  'nonce_str': self.param_dict['nonce_str'],
                  'filter': str(filter),
                  'session_id': str(random.randint(1, 9999999))
                  }
        self.param_dict['image'] = base64_data
        sign_before = ''
        # 要对key排序再拼接
        for key in sorted(params):
            # 键值拼接过程value部分需要URL编码，URL编码算法用大写字母，例如%E8。quote默认大写。
            sign_before += '{}={}&'.format(key, quote(params[key], safe=''))
        # 将应用密钥以app_key为键名，拼接到字符串sign_before末尾
        sign_before += 'app_key={}'.format(self.param_dict['app_key'])
        '''计算MD5摘要，得到签名字符串'''
        sign = hashlib.md5(sign_before.encode('utf-8')).hexdigest()
        sign = sign.upper()
        params['sign'] = sign
        return params

    def get_content(self, image, filter):
        params = self.get_params(image, filter)
        try:
            r = requests.post(self.interface_url, data=params)
            soup = BeautifulSoup(r.text, 'lxml')
            allcontents = soup.select('body')[0].text.strip()
            allcontents_json = json.loads(allcontents)  # str转成dict
            image_data = base64.b64decode(allcontents_json["data"]['image'])
            with open("filtered_image.jpg", 'wb') as f:
                f.write(image_data)
        except:
            print('exception occur')

if __name__ == '__main__':
    img_filter = VisionImgFilter('2118177424', 'dTBQojzFDaabB0A2')
    with open('D:/ico/mmexport1547432113759.jpg', 'rb')as f:
        img_filter.get_content(f.read(), 1)

原始图像：

处理后：

ocr_generalocr.py

'''
date:2019.7.7
copyright:buaalzm
'''
from qqaibase import QQAIBase
import hashlib
import requests
from bs4 import BeautifulSoup
import json
from urllib.parse import quote
import base64


class GeneralOCR(QQAIBase):
    interface_url = "https://api.ai.qq.com/fcgi-bin/ocr/ocr_generalocr"  # API地址
    def get_params(self, image):
        self.basic_param_init()
        '''图片用base64编码'''
        base64_data = base64.b64encode(image)
        params = {'app_id': self.param_dict['app_id'],
                  'image': base64_data,
                  'time_stamp': self.param_dict['time_stamp'],
                  'nonce_str': self.param_dict['nonce_str'],
                  }
        self.param_dict['image'] = base64_data
        sign_before = ''
        # 要对key排序再拼接
        for key in sorted(params):
            # 键值拼接过程value部分需要URL编码，URL编码算法用大写字母，例如%E8。quote默认大写。
            sign_before += '{}={}&'.format(key, quote(params[key], safe=''))
        # 将应用密钥以app_key为键名，拼接到字符串sign_before末尾
        sign_before += 'app_key={}'.format(self.param_dict['app_key'])
        '''计算MD5摘要，得到签名字符串'''
        sign = hashlib.md5(sign_before.encode('utf-8')).hexdigest()
        sign = sign.upper()
        params['sign'] = sign
        return params

    def get_content(self, image):
        params = self.get_params(image)
        try:
            r = requests.post(self.interface_url, data=params)
            soup = BeautifulSoup(r.text, 'lxml')
            allcontents = soup.select('body')[0].text.strip()
            allcontents_json = json.loads(allcontents)  # str转成dict
            return allcontents_json["data"]
        except:
            print('exception occur')

if __name__ == '__main__':
    img_filter = GeneralOCR('2118233756', 'YSh9DEi7NNIH8Eiq')
    with open('1.jpg', 'rb')as f:
        image_data = f.read()
        for item in img_filter.get_content(image_data)['item_list']:
            print(item['itemstring'])

原始图像：

输出结果：