使用腾讯OCR，图片转表格

不是广告哈，就是记录一下操作，因为这些页面真的很难找，即便知道缺什么也找不到对应位置在哪…

调用说明:（以Python为例）

整个流程：

1. 腾讯开服务
2. 图片转base64编码
3. 把base64编码转腾讯OCR
4. 腾讯OCR返回的json，转Pandas Dict类型
5. 整理dict的数据为Pandas DataFrame类型

步骤1：腾讯开服务

1. 进入控制台：https://console.cloud.tencent.com/cam/capi，新建一个密钥：
   

2. 在控制台开启资源包：https://console.cloud.tencent.com/ocr/overview

步骤2：图片转Base64使用

def change_img_to_base64(image_path):
    """base64编码图片"""
    with open(image_path, 'rb') as f:
        image_data = f.read()
        base64_data: bytes = base64.b64encode(image_data)  # base64编码
        return base64_data

步骤3：把Base64编码转腾讯OCR

def tencent_ocr(suffix, image_based_64):
    """腾讯OCR
    :param suffix:图片的后缀，比如png,jpg
    :param image_based_64:图片的base64编码
    """
    from tencentcloud.common import credential
    from tencentcloud.common.profile.client_profile import ClientProfile
    from tencentcloud.common.profile.http_profile import HttpProfile
    from tencentcloud.common.exception.tencent_cloud_sdk_exception import TencentCloudSDKException
    from tencentcloud.ocr.v20181119 import ocr_client, models
    try:
        cred = credential.Credential("自己的SecretId(看步骤1)", "自己的SecretKey(看步骤1)")
        httpProfile = HttpProfile()
        httpProfile.endpoint = "ocr.tencentcloudapi.com"

        clientProfile = ClientProfile()
        clientProfile.httpProfile = httpProfile
        client = ocr_client.OcrClient(cred, "ap-beijing", clientProfile)

        req = models.RecognizeTableOCRRequest()
        params = {
            "ImageBase64": "data:image/{suffix};base64,{image_based_64}".format(
                suffix=suffix, image_based_64=image_based_64.decode("utf8"))
        }
        req.from_json_string(json.dumps(params))
        resp = client.RecognizeTableOCR(req)
        return resp.to_json_string()
    except TencentCloudSDKException as err:
        print(err)

步骤4：Json转Dict

def json_to_dict(my_json: str) -> dict:
    """json转dict类型"""
    return json.loads(my_json)

步骤5：Dict数据转Pandas.DataFrame格式：

def formation(json_data):
    """根据腾讯ocr识别结果整理格式并输出"""
    if json_data is not None:
        dict_data = json_to_dict(json_data)
        for table_index, table_data in enumerate(dict_data['TableDetections']):  # 遍历每个表格的数据
            table_data_list = []
            for each_table_data in table_data['Cells']:  # 遍历每个表中的数据
                content: str = each_table_data['Text']
                x_y: list = list(each_table_data['Polygon'][0].values())
                x_y.append(content)
                table_data_list.append(x_y)
            # 整理格式
            table_df = pd.DataFrame(table_data_list)
            table_df.columns = ['x', 'y', 'content']
            table_df.sort_values(by=['y', 'x'], ascending=True, inplace=True)
            for index, line_df in table_df.groupby("y"):  # 按每行进行处理
                line_df.sort_values(by=['x'], ascending=True, inplace=True)
                line_values = line_df['content'].values  # 当前行的数据
                print(",".join(line_values))

步骤6：整体依次调用

if __name__ == '__main__':
    image = "mypic.png"
    image_base64 = change_img_to_base64(image)  # 步骤1：图片转base64
    suffix = image.split('.')[-1]  # 后缀
    tencent_result: json = tencent_ocr(suffix, image_base64)
    formation(tencent_result)

完整代码

import base64
import json
import pandas as pd


def change_img_to_base64(image_path):
    """base64编码图片"""
    with open(image_path, 'rb') as f:
        image_data = f.read()
        base64_data: bytes = base64.b64encode(image_data)  # base64编码
        return base64_data


def json_to_dict(my_json: str) -> dict:
    """json转dict类型"""
    return json.loads(my_json)


def tencent_ocr(suffix, image_based_64):
    """腾讯OCR
    :param suffix:图片的后缀，比如png,jpg
    :param image_based_64:图片的base64编码
    """
    from tencentcloud.common import credential
    from tencentcloud.common.profile.client_profile import ClientProfile
    from tencentcloud.common.profile.http_profile import HttpProfile
    from tencentcloud.common.exception.tencent_cloud_sdk_exception import TencentCloudSDKException
    from tencentcloud.ocr.v20181119 import ocr_client, models
    try:
        cred = credential.Credential("自己的SecretId(看步骤1)", "自己的SecretKey(看步骤1)")
        httpProfile = HttpProfile()
        httpProfile.endpoint = "ocr.tencentcloudapi.com"

        clientProfile = ClientProfile()
        clientProfile.httpProfile = httpProfile
        client = ocr_client.OcrClient(cred, "ap-beijing", clientProfile)

        req = models.RecognizeTableOCRRequest()
        params = {
            "ImageBase64": "data:image/{suffix};base64,{image_based_64}".format(
                suffix=suffix, image_based_64=image_based_64.decode("utf8"))
        }
        req.from_json_string(json.dumps(params))
        resp = client.RecognizeTableOCR(req)
        return resp.to_json_string()
    except TencentCloudSDKException as err:
        print(err)


def formation(json_data):
    """根据腾讯ocr识别结果整理格式并输出"""
    if json_data is not None:
        dict_data = json_to_dict(json_data)
        for table_index, table_data in enumerate(dict_data['TableDetections']):  # 遍历每个表格的数据
            table_data_list = []
            for each_table_data in table_data['Cells']:  # 遍历每个表中的数据
                content: str = each_table_data['Text']
                x_y: list = list(each_table_data['Polygon'][0].values())
                x_y.append(content)
                table_data_list.append(x_y)
            # 整理格式
            table_df = pd.DataFrame(table_data_list)
            table_df.columns = ['x', 'y', 'content']
            table_df.sort_values(by=['y', 'x'], ascending=True, inplace=True)
            for index, line_df in table_df.groupby("y"):  # 按每行进行处理
                line_df.sort_values(by=['x'], ascending=True, inplace=True)
                line_values = line_df['content'].values  # 当前行的数据
                print(",".join(line_values))


if __name__ == '__main__':
    image = "mypic.png"
    image_base64 = change_img_to_base64(image)  # 步骤1：图片转base64
    suffix = image.split('.')[-1]  # 后缀
    tencent_result: json = tencent_ocr(suffix, image_base64)
    formation(tencent_result)

资料整理

• 腾讯提供的服务叫做行业文档识别：https://cloud.tencent.com/act/event/ocrdemo?dt=4

• 行业文档识别的接口文档：https://cloud.tencent.com/document/api/866/49525

• API调试的界面在：https://console.cloud.tencent.com/api/explorer?Product=ocr&Version=2018-11-19&Action=RecognizeTableOCR&SignVersion=