Skip to content

fengzhizi715/Tess-TwoDemo

Repository files navigation

Tess-TwoDemo

tess-two是Tesseract在Android平台上的移植。

下载tess-two:

compile 'com.rmtheis:tess-two:8.0.0'

然后将训练好的eng.traineddata放入android项目的assets文件夹中,就可以识别英文了。

1. 简单地识别英文

初始化tess-two,加载训练好的tessdata

    private void prepareTesseract() {
        try {
            prepareDirectory(DATA_PATH + TESSDATA);
        } catch (Exception e) {
            e.printStackTrace();
        }

        copyTessDataFiles(TESSDATA);
    }

    /**
     * Prepare directory on external storage
     *
     * @param path
     * @throws Exception
     */
    private void prepareDirectory(String path) {

        File dir = new File(path);
        if (!dir.exists()) {
            if (!dir.mkdirs()) {
                Log.e(TAG, "ERROR: Creation of directory " + path + " failed, check does Android Manifest have permission to write to external storage.");
            }
        } else {
            Log.i(TAG, "Created directory " + path);
        }
    }

    /**
     * Copy tessdata files (located on assets/tessdata) to destination directory
     *
     * @param path - name of directory with .traineddata files
     */
    private void copyTessDataFiles(String path) {
        try {
            String fileList[] = getAssets().list(path);

            for (String fileName : fileList) {

                // open file within the assets folder
                // if it is not already there copy it to the sdcard
                String pathToDataFile = DATA_PATH + path + "/" + fileName;
                if (!(new File(pathToDataFile)).exists()) {

                    InputStream in = getAssets().open(path + "/" + fileName);

                    OutputStream out = new FileOutputStream(pathToDataFile);

                    // Transfer bytes from in to out
                    byte[] buf = new byte[1024];
                    int len;

                    while ((len = in.read(buf)) > 0) {
                        out.write(buf, 0, len);
                    }
                    in.close();
                    out.close();

                    Log.d(TAG, "Copied " + fileName + "to tessdata");
                }
            }
        } catch (IOException e) {
            Log.e(TAG, "Unable to copy files to tessdata " + e.toString());
        }
    }

取景框.JPG

拍完照后,调用startOCR方法。

    private void startOCR(Uri imgUri) {
        try {
            BitmapFactory.Options options = new BitmapFactory.Options();
            options.inSampleSize = 4; // 1 - means max size. 4 - means maxsize/4 size. Don't use value <4, because you need more memory in the heap to store your data.
            Bitmap bitmap = BitmapFactory.decodeFile(imgUri.getPath(), options);

            String result = extractText(bitmap);
            resultView.setText(result);

        } catch (Exception e) {
            Log.e(TAG, e.getMessage());
        }
    }

extractText()会调用tess-two的api来实现ocr文字识别。

    private String extractText(Bitmap bitmap) {
        try {
            tessBaseApi = new TessBaseAPI();
        } catch (Exception e) {
            Log.e(TAG, e.getMessage());
            if (tessBaseApi == null) {
                Log.e(TAG, "TessBaseAPI is null. TessFactory not returning tess object.");
            }
        }

        tessBaseApi.init(DATA_PATH, lang);

        tessBaseApi.setImage(bitmap);
        String extractedText = "empty result";
        try {
            extractedText = tessBaseApi.getUTF8Text();
        } catch (Exception e) {
            Log.e(TAG, "Error in recognizing text.");
        }
        tessBaseApi.end();
        return extractedText;
    }

最后,显示识别的效果,此时的效果还算可以。 简单地识别英文.JPG

2. 识别代码

接下来,尝试用上面的程序识别一段代码。 识别代码.JPG

此时,效果一塌糊涂。我们重构一下startOCR(),增加局部的二值化处理。

    private void startOCR(Uri imgUri) {
        try {
            BitmapFactory.Options options = new BitmapFactory.Options();
            options.inSampleSize = 4; // 1 - means max size. 4 - means maxsize/4 size. Don't use value <4, because you need more memory in the heap to store your data.
            Bitmap bitmap = BitmapFactory.decodeFile(imgUri.getPath(), options);

            CV4JImage cv4JImage = new CV4JImage(bitmap);
            Threshold threshold = new Threshold();
            threshold.adaptiveThresh((ByteProcessor)(cv4JImage.convert2Gray().getProcessor()), Threshold.ADAPTIVE_C_MEANS_THRESH, 12, 30, Threshold.METHOD_THRESH_BINARY);
            Bitmap newBitmap = cv4JImage.getProcessor().getImage().toBitmap(Bitmap.Config.ARGB_8888);

            ivImage2.setImageBitmap(newBitmap);

            String result = extractText(newBitmap);
            resultView.setText(result);

        } catch (Exception e) {
            Log.e(TAG, e.getMessage());
        }
    }

在这里,使用cv4j来实现图像的二值化处理。

            CV4JImage cv4JImage = new CV4JImage(bitmap);
            Threshold threshold = new Threshold();
            threshold.adaptiveThresh((ByteProcessor)(cv4JImage.convert2Gray().getProcessor()), Threshold.ADAPTIVE_C_MEANS_THRESH, 12, 30, Threshold.METHOD_THRESH_BINARY);
            Bitmap newBitmap = cv4JImage.getProcessor().getImage().toBitmap(Bitmap.Config.ARGB_8888);

图像二值化就是将图像上的像素点的灰度值设置为0或255,也就是将整个图像呈现出明显的黑白效果。图像的二值化有利于图像的进一步处理,使图像变得简单,而且数据量减小,能凸显出感兴趣的目标的轮廓。

cv4j的github地址:https://github.com/imageprocessor/cv4j

cv4j 是gloomyfish和我一起开发的图像处理库,纯java实现。

再来试试效果,图片中间部分是二值化后的效果,此时基本能识别出代码的内容。 先做二值化再识别代码.JPG

3. 识别中文

如果要识别中文字体,需要使用中文的数据包。可以去下面的网站上下载。

https://github.com/tesseract-ocr/tessdata

跟中文相关的数据包有chi_sim.traineddata、chi_tra.traineddata,它们分别表示是简体中文和繁体中文。

tessBaseApi.init(DATA_PATH, lang);

前面的例子都是识别英文的,所以原先的lang值为"eng",现在要识别简体中文的话需要将其值改为"chi_sim"。

识别中文.JPG

最后

本项目只是demo级别的演示,离生产环境的使用还差的很远。 本项目的github地址:https://github.com/fengzhizi715/Tess-TwoDemo

为何说只是demo级别呢?

  • 数据包很大,特别是中文的大概有50多M,放在移动端的肯定不合适。一般正确的做法,都是放在云端。
  • 识别文字很慢,特别是中文,工程上还有很多优化的空间。
  • 做ocr之前需要做很多预处理的工作,在本例子中只用了二值化,其实还有很多预处理的步骤比如倾斜校正、字符切割等等。
  • 为了提高tess-two的识别率,可以自己训练数据集。

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages