Erlang/Elixir: 使用 OpenCV, Python 搭建图片缩略图服务器

这篇文章是在OSX上测试和运行的的, Ubuntu下的安装和配置请移步到这里

应用程序进程树, 默认启动 Poolboy 时, 初始化10个用于处理图片的 Python 工作进程(Worker)

首先安装OpenCV需要的工具包

ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
brew install python
brew tap homebrew/science
brew install opencv
sudo pip install numpy
sudo pip install matplotlib

使用 Homebrew 的 Python 版本, 而不是 Mac OS X 系统自带的 Python

alias python='/usr/local/bin/python'

创建 Elixir 项目

➜  mix new opencv_thumbnail_server --sup
* creating README.md
* creating .gitignore
* creating mix.exs
* creating config
* creating config/config.exs
* creating lib
* creating lib/opencv_thumbnail_server.ex
* creating test
* creating test/test_helper.exs
* creating test/opencv_thumbnail_server_test.exs

Your Mix project was created successfully.
You can use "mix" to compile it, test it, and more:

    cd opencv_thumbnail_server
    mix test

Run "mix help" for more commands.

Elixir 模块

require Logger
defmodule OpencvThumbnailServer do
  use Application
  def start(_type, _args) do
    Logger.info "Start opencv thumbnail server"
    OpencvThumbnailServer.Supervisor.start_link()
  end
end

defmodule OpencvThumbnailServer.Supervisor do
  use Supervisor

  @config Application.get_env :opencv_thumbnail_server, :settings

  def start_link() do
    Supervisor.start_link(__MODULE__, [], name: {:global,__MODULE__})
  end

  def init([]) do
    pool_options = @config[:poolboy]
    {_, name} = pool_options[:name]
    children = [
      :poolboy.child_spec(name, pool_options, @config[:module_name])
    ]
    supervise(children, strategy: :one_for_all, max_restarts: 1000, max_seconds: 3600)
  end
end

defmodule OpencvThumbnailServer.Worker do
  use GenServer
  @config Application.get_env(:opencv_thumbnail_server, :settings)

  def start_link(python_module) do
    GenServer.start_link(__MODULE__, python_module, [])
  end

  def call_python(worker, function, args) do
    GenServer.call(worker, {:call_python, function, args}, 10_000)
  end

  def init(python_module) do
    IO.puts "Start worker"
    {:ok, pid} = :python.start_link([
      {:python_path, @config[:python_path]},
      {:python, @config[:python]}
    ])
    state = {python_module, pid}
    {:ok, state}
  end

  def handle_call({:call_python, function, args}, _from, state) do
    {module, pid} = state
    result = :python.call(pid, module, function, args)
    reply = {:ok, result}
    {:reply, reply, state}
  end

  def handle_call(_request, _from, state) do
    {:stop, :error, :bad_call, state}
  end

  def handle_info(_msg, {module,py_pid}) do
    {:stop, :error, {module,py_pid}}
  end

  def terminate(_reason, {_, py_pid}) do
    :python.stop(py_pid)
    :ok
  end
end

图像处理

获取宽高

# -*- coding: utf-8 -*-

import urllib2 as urllib
import numpy as np
import cv2


def load_image_url(url):
    resp = urllib.urlopen(url)
    buf = resp.read()
    return buf


def load_image_file(filename):
    image = cv2.imdecode(filename, cv2.IMREAD_COLOR)
    return image

def get_photo_sizes():
    return [
        [160, 160],
        [320, 320],
        [640, 640],
        [1060, 1060],
        [1280, 1280]
    ]
def show(buf):
    # print buf
    # x = cv2.imdecode(image, cv2.IMREAD_COLOR)
    # d = cv2.cvtColor(c, cv2.COLOR_RGB2BGR)
    np_ndarray = np.fromstring(buf, dtype=np.uint8)
    x = cv2.imdecode(np_ndarray, cv2.IMREAD_UNCHANGED)
    return cv2.imshow('NBA Image', x)

def write(buf):
    nparray = np.fromstring(buf, dtype=np.uint8)
    img = cv2.imdecode(nparray, cv2.IMREAD_UNCHANGED)
    return cv2.imwrite('/tmp/imwrite.png', img)

# def get_dimension():
#     url = 'http://img1.gtimg.com/16/1601/160106/16010642_1200x1000_0.jpg'
#     resp = urllib.urlopen(url)
#     buf = resp.read()
#     x = np.fromstring(buf, dtype=np.uint8)
#     img = cv2.imdecode(x, cv2.IMREAD_UNCHANGED)
#     # height = np.size(img, 0)
#     # width = np.size(img, 1)
#     height, width = image.shape[:2]
#     return (width, height)

def get_dimension(buffer):
    # 把原始的二进制图片数据转换为NpArray
    nparray = np.fromstring(buffer, dtype=np.uint8)
    # 把 nparray 转换为 opencv 的图像格式
    image = cv2.imdecode(nparray, cv2.IMREAD_UNCHANGED)
    height, width = image.shape[:2]
    return (width, height)

def convert_color():
    url = 'http://ww3.sinaimg.cn/mw690/6941baebgw1epzcuv9vmxj20me0hy0u1.jpg'
    resp = urllib.urlopen(url)
    buf = resp.read()
    x = np.fromstring(buf, dtype=np.uint8)
    img = cv2.imdecode(x, cv2.IMREAD_UNCHANGED)

if __name__ == '__main__':
    get_dimension()

在 Erlang 和 Python 之间传输二进制数据

Erlang 的binary()数据类型和 Python 之间的映射关系, 在Python 2.x 中二进制数据类型为 str() 表示, Python 3.x 中为 bytes()

buf = resp.read(), 其中变量 buf 的类型为 <type 'str'>

在 Elixir 我们看到变量 buf 的值为:

{:ok, <<255, 216, 255, 224, 0, 16, 74, 70, 73, 70, 0, 1, 1, 1, 0, 72, 
      0, 72, 0, 0, 255, 219, 0, 67, 0, 8, 6, 6, 7, 6, 5, 8, 7, 7, 7, 
      9, 9, 8, 10, 12, 20, 13, 12, 11, 11, 12, 25, 18, 19, ...>>}

调用 Python 函数

{:ok, data} = OpencvThumbnailServer.Api.load_image_url("https://segmentfault.com/img/bVwhAW")
OpencvThumbnailServer.Api.get_dimension(data) 
{:ok, {800, 431}}

创建 Python 模块

之前的 Python 图像处理模块可以组织到一个项目中单独维护. 这里使用工具 cookiecutter 创建 Python 一个基本的项目骨架, 用于实现缩略图的功能

cookiecutter 可以通过多种方式安装, 包括pip, easy_install, conda, brew

pip install cookiecutter
easy_install cookiecutter
conda install -c https://conda.binstar.org/pydanny cookiecutter
brew install cookiecutter(Mac OS X)

目录结构

➜  opencv_thumbnail git:(master) tree
.
├── AUTHORS.rst
├── CONTRIBUTING.rst
├── HISTORY.rst
├── LICENSE
├── MANIFEST.in
├── Makefile
├── README.rst
├── build
│   ├── bdist.macosx-10.11-x86_64
│   └── lib
│       └── opencv_thumbnail
│           ├── __init__.py
│           └── opencv_thumbnail.py
├── dist
│   └── opencv_thumbnail-0.1.0-py2.7.egg
├── docs
│   ├── Makefile
│   ├── authors.rst
│   ├── conf.py
│   ├── contributing.rst
│   ├── history.rst
│   ├── index.rst
│   ├── installation.rst
│   ├── make.bat
│   ├── readme.rst
│   └── usage.rst
├── opencv_thumbnail
│   ├── __init__.py
│   ├── __init__.pyc
│   ├── opencv_thumbnail.py
│   └── opencv_thumbnail.pyc
├── opencv_thumbnail.egg-info
│   ├── PKG-INFO
│   ├── SOURCES.txt
│   ├── dependency_links.txt
│   ├── not-zip-safe
│   └── top_level.txt
├── requirements_dev.txt
├── setup.cfg
├── setup.py
├── tests
│   ├── __init__.py
│   └── test_opencv_thumbnail.py
├── tox.ini
└── travis_pypi_setup.py

9 directories, 36 files

API实现

调用需要从 Poolboy 池中取出一个工作进程, 并调用工作进程的call_python, 进程使用完成后返还给 Poolboy 进程池, 这里对调用过程封装一下, 以简化使用.

defmodule OpencvThumbnailServer.Api do
  alias OpencvThumbnailServer.Worker

  def get_dimension(data) do
    worker = :poolboy.checkout(:opencv_thumbnail_server_pool)
    {w, h} = Worker.call_python(worker, :get_dimension, [data])
    :poolboy.checkin(:opencv_thumbnail_server_pool, worker)
    {w, h}
  end

  def load_image_url(url) do
    worker = :poolboy.checkout(:opencv_thumbnail_server_pool)
    image_bin = Worker.call_python(worker, :load_image_url, [url])
    :poolboy.checkin(:opencv_thumbnail_server_pool, worker)
    image_bin
  end
end

源码

https://github.com/developerworks/opencv_thumbnail_server

参考资料

利用Python和OpenCV将URL直接转换成OpenCV格式
How to read raw png from an array in python opencv?
Install OpenCV for Python on Mac OS X
Installing scikit-image
How can i read an image from an internet url in python cv2 , scikit image and mahotas
Using Elixir, erlport with Python 2.7.9, receiving an arity error
How to read image from in memory buffer (StringIO) or from url with opencv python library
Python OpenCV convert image to byte string?