很多外包公司都有一套自动化测试平台，可以销售给甲方（银行保险之类公司）；而像一些互联网大厂通过自己技术人员进行开发，一般实在appium基础上再进行深度开发。AI自动化思路是结合appium加入图像识别以及ocr，在八大定位元素基础上，增加图像识别以及文本识别（ocr），实现自动化测试。
以下是工作app中登录实战，使用了百度飞浆进行ocr识别，然后计算出坐标位置，进行点击操作，然后再输入用户名
from appium import webdriver
import time
from config import setting
from paddleocr import PaddleOCR
desired_caps = dict()
desired_caps['platformName'] = 'Android' # 可以写成android
desired_caps['platformVersion'] = '12' # 11.1.0等都可以写成11
desired_caps['deviceName'] = '测试手机型号' # 设备名字可以随便写，但是不可以为空
desired_caps['appPackage'] = '测试app包'
desired_caps['appActivity'] = '你测试app的MainActivity'
desired_caps['noReset'] = True # 打开app时不清除数据
desired_caps['udid'] = "你的测试设备"
# desired_caps['udid'] = "80fe4bfb"
# desired_caps['udid'] = "9WDNW21107036993"
# desired_caps['automationName'] = "UiAutomator2"
desired_caps['unicodeKeyboard'] = True
desired_caps['resetKeyboard'] = True
#启动appium服务
# cmd = r'start appium -a 127.0.0.1 -p 4723 -bp 4724'
# pr = subprocess.Popen(cmd, stdout=subprocess.PIPE, shell=True)
# pr.wait(timeout=3)
driver = webdriver.Remote('http://localhost:4723/wd/hub', desired_caps)
time.sleep(10) # 等待20秒"
img_folder = setting.OCR_DIR
screen_save_path = img_folder + '登录.png'
# # time1 = time.strftime('%Y%m%d%H%M', time.localtime(time.time()))
# # screen_save_path = img_folder + time1 + '.png'
driver.get_screenshot_as_file(screen_save_path)
ocr = PaddleOCR()
result = ocr.ocr(screen_save_path)
point = []
print(result)
for line in result[0]:
if line[1][0] == '用户名':
print(line[0])
# left, top, right, bottom = line[0]
# print(top)
a = 0
b = 0
for line1 in line[0]:
a = a + line1[0]
b = b + line1[1]
# print(a)
# print(b)
c = a // 4
e = b // 4
point.append(c)
point.append(e)
x, y = point
time.sleep(2)
# action.tap(x=x, y=y).perform()
driver.tap([(x, y)], 100)
time.sleep(3)
driver.switch_to.active_element.send_keys("编程老怪") # 输入文本
time.sleep(20)
driver.quit()
