如何识别转换pdf文档

pdf转图片

package main

//sudo apt install libmagic-dev libmagickwand-dev

import (
	"strconv"
	"gopkg.in/gographics/imagick.v2/imagick"
)

func ConvertPdfToJpg(pdfName string, imageName string, pdfSize int) error {

	imagick.Initialize()
	defer imagick.Terminate()

	mw := imagick.NewMagickWand()
	defer mw.Destroy()
	
	if err := mw.SetResolution(200, 200); err != nil {
		return err
	}
	
	if err := mw.ReadImage(pdfName); err != nil {
		return err
	}

	if err := mw.SetImageAlphaChannel(imagick.ALPHA_CHANNEL_FLATTEN); err != nil {
		return err
	}

	if err := mw.SetCompressionQuality(90); err != nil {
		return err
	}
	
	if err := mw.SetFormat("jpg"); err != nil {
		return err
	}
	for i := 0; i < pdfSize; i++ {
		mw.SetIteratorIndex(i)
		if err := mw.WriteImage(imageName + "_" + strconv.Itoa(i) + ".jpg"); err != nil {
			return err
		}
	}
	return nil
}

图片识别

原理可参考：图像识别

tesseract-ocr

tesseract-ocr

tesseract test_0.jpg output --oem 1 -l chi_sim+eng

you can use command, develop api,also you can try tranning

百度云OCR

百度文字识别OCR

个人有需求：偶尔需要识别pdf图片中的文字形成文本文档，为方便使用最好能生成可执行文件点击就可用，尽量减少依赖，故这里使用golang调用。

百度AI平台QuickStart文档配套DEMO程序的python实现

阿里，腾讯等云平台也有相关服务，替换下面代码中的请求api即可。

1. 登录云平台创建授权应用(一般是oauth2机制)；
2. 打开对应平台的OCR接口说明，百度通用文字识别(标准版)
3. 实现代码如下：

package main

import (
	"bufio"
	"encoding/base64"
	"encoding/json"
	"fmt"
	"io/ioutil"
	"log"
	"net/http"
	"net/url"
	"os"
	"strings"
	"time"
)

var maxLen int = 5 * 1024 * 1024

var API_KEY string = ""

var SECRET_KEY string = ""

var OCR_URL string = "https://aip.baidubce.com/rest/2.0/ocr/v1/general_basic?access_token="

var TOKEN_URL string = "https://aip.baidubce.com/oauth/2.0/token?grant_type=client_credentials"

// 通用文字识别（标准版） https://ai.baidu.com/ai-doc/OCR/zk3h7xz52
func GeneralBasic(name string, access_token string) {

	if access_token == "" {
		log.Fatal("access_token is empty")
	}

	fmt.Println("OCR File[" + name + "]")

	file, _ := os.Open(name)
	defer file.Close()

	time.Sleep(2 * time.Second) //qps limit 1

	OCR_URL += access_token
	res, err := http.Post(OCR_URL, "application/x-www-form-urlencoded", strings.NewReader("image="+url.QueryEscape(base64e(file))))
	if err != nil {
		log.Fatal(err)
	}
	defer res.Body.Close()
	mapdata := make(map[string]interface{})
	json.NewDecoder(res.Body).Decode(&mapdata)
	if mapdata["words_result_num"] == nil {
		data, _ := ioutil.ReadAll(res.Body)
		fmt.Printf("%s identification failed:%+v\n", file.Name(), data)
		return
	}
	if int(mapdata["words_result_num"].(float64)) == 0 {
		fmt.Println(file.Name() + " identification content is empty...")
		return
	}

	f, err := os.OpenFile(file.Name()+".txt", os.O_RDWR|os.O_CREATE, os.ModePerm)
	if err != nil {
		log.Fatal(err)
	}
	defer f.Close()

	w := bufio.NewWriter(f)
	words_result := mapdata["words_result"].([]interface{})

	for _, obj := range words_result {
		m := obj.(map[string]interface{})
		fmt.Fprintln(w, m["words"])
	}
	w.Flush()
}

// 鉴权认证机制 http://ai.baidu.com/ai-doc/REFERENCE/Ck3dwjhhu
// post方式报错:unsupported_grant_type
func Fetch_token() string {
	res, err := http.Get(TOKEN_URL + "&client_id=" + API_KEY + "&client_secret=" + SECRET_KEY)
	if err != nil {
		log.Fatal(err)
	}
	defer res.Body.Close()
	// data, err := ioutil.ReadAll(res.Body)
	if err != nil {
		log.Fatal(err)
	}
	mapdata := make(map[string]interface{})
	json.NewDecoder(res.Body).Decode(&mapdata)
	// json.Unmarshal(data, &mapdata)
	for key, value := range mapdata {
		// fmt.Println("key:", key, " => value :", value)
		if key == "access_token" {
			return value.(string)
		}
		if key == "error_description" {
			log.Fatal(value)
		}
	}
	return ""
}

// base64 encode
func base64e(file *os.File) string {
	sourcebuffer := make([]byte, maxLen)
	n, _ := file.Read(sourcebuffer)
	return base64.StdEncoding.EncodeToString(sourcebuffer[:n])
}

// base64 decode
func base64d(str string) ([]byte, error) {
	return base64.StdEncoding.DecodeString(str)
}

github上有测试样例，替换地址及添加自己的api_key和secret_key即可