Go语言中OCR常用识别库的使用与实战指南

sudo apt update
sudo apt install -y tesseract-ocr tesseract-ocr-eng
# 需要其他语言：tesseract-ocr-chi-sim / tesseract-ocr-chi-tra / tesseract-ocr-fra ...

brew install tesseract
brew install tesseract-lang # 或手动安装对应的 traineddata

go get github.com/otiai10/gosseract/v2

go get github.com/otiai10/gosseract/v2
go get gocv.io/x/gocv
go get github.com/disintegration/imaging
go get github.com/go-resty/resty/v2    # 如果需要从 URL 下载图片

package main

import (
	"fmt"
	"image"
	"image/color"
	"io"
	"log"
	"net/http"
	"os"

	"github.com/disintegration/imaging"
	"github.com/go-resty/resty/v2"
	"github.com/otiai10/gosseract/v2"
	"gocv.io/x/gocv"
)

// 下载图片到本地临时文件
func downloadImage(url string) (string, error) {
	client := resty.New()
	resp, err := client.R().SetDoNotParseResponse(true).Get(url)
	if err != nil {
		return "", err
	}
	defer resp.RawBody().Close()

	tmp := "tmp_download.jpg"
	out, err := os.Create(tmp)
	if err != nil {
		return "", err
	}
	defer out.Close()

	_, err = io.Copy(out, resp.RawBody())
	if err != nil {
		return "", err
	}
	return tmp, nil
}

// 使用 gocv 对图片做预处理，返回处理后保存的文件路径
func preprocessWithGocv(srcPath string) (string, error) {
	img := gocv.IMRead(srcPath, gocv.IMReadColor)
	if img.Empty() {
		return "", fmt.Errorf("image empty")
	}
	defer img.Close()

	// resize: 保证短边 >= 800 提升 OCR 精度（可按需调整）
	h, w := img.Rows(), img.Cols()
	scale := 1.0
	minSide := w
	if h < w {
		minSide = h
	}
	if minSide < 800 {
		scale = 800.0 / float64(minSide)
		gocv.Resize(img, &img, image.Point{}, scale, scale, gocv.InterpolationLinear)
	}

	// 转灰度
	gocv.CvtColor(img, &img, gocv.ColorBGRToGray)

	// 高斯模糊 去噪
	gocv.GaussianBlur(img, &img, image.Point{X: 3, Y: 3}, 0, 0, gocv.BorderDefault)

	// 自适应阈值
	gocv.AdaptiveThreshold(img, &img, 255, gocv.AdaptiveThresholdGaussian, gocv.ThresholdBinaryInv, 25, 15)

	// 形态学开闭操作（去小斑点、连接文字）
	kernel := gocv.GetStructuringElement(gocv.MorphRect, image.Pt(3, 3))
	defer kernel.Close()
	gocv.MorphologyEx(img, &img, gocv.MorphClose, kernel)

	outPath := "tmp_preprocessed.png"
	gocv.IMWrite(outPath, img)
	return outPath, nil
}

// 用 imaging 做简单预处理（替代 gocv 的简单流程）
func preprocessWithImaging(srcPath string) (string, error) {
	img, err := imaging.Open(srcPath)
	if err != nil {
		return "", err
	}
	// 灰度
	gray := imaging.Grayscale(img)
	// 锐化（可选）
	gray = imaging.Sharpen(gray, 1.0)
	outPath := "tmp_pre_img.png"
	err = imaging.Save(gray, outPath)
	if err != nil {
		return "", err
	}
	return outPath, nil
}

// 调用 gosseract 识别图片
func doOCR(imagePath string, lang string) (string, error) {
	client := gosseract.NewClient()
	defer client.Close()

	// 设置 tessdata path in case 非默认安装路径（可选）
	// client.SetTessdataPrefix("/usr/share/tessdata")

	if lang != "" {
		client.SetLanguage(lang) // 常见: "eng", "chi_sim", "chi_tra", "fra", ...
	}

	// 可设置识别参数来提升速度或精度
	client.SetPageSegMode(gosseract.PSM_AUTO) // 或 PSM_SINGLE_BLOCK 等
	// 字符白名单示例（只识别数字）
	// client.SetVariable("tessedit_char_whitelist", "0123456789")

	client.SetImage(imagePath)
	text, err := client.Text()
	if err != nil {
		return "", err
	}
	return text, nil
}

func main() {
	// 示例：从 URL 下载 -> 预处理 -> OCR
	url := "https://example.com/sample.jpg"
	src, err := downloadImage(url)
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println("downloaded:", src)

	// 选择用 gocv 还是 imaging
	prep, err := preprocessWithGocv(src)
	if err != nil {
		// fallback
		prep, err = preprocessWithImaging(src)
		if err != nil {
			log.Fatal(err)
		}
	}
	fmt.Println("preprocessed:", prep)

	// 多语言示例：中文识别
	out, err := doOCR(prep, "chi_sim")
	if err != nil {
		log.Fatal(err)
	}
	fmt.Println("OCR result:\n", out)
}