使用Firebase ML Kit文字识别API遇报错，转图片格式后仍异常

Let's tackle the two crash scenarios you're facing with Firebase ML Kit Text Recognition, and walk through actionable fixes for each.

First Crash (JPG Format) #

Your initial error shows ML Kit failing to decode pixel data from the JPG:

A/native: image_util_common.cc:2757 Check failed: 'out_pix' Must be non NULL terminating.
A/native: image_util_common.cc:2797 Check failed: 'out_pix' Must be non NULL
A/libc: Fatal signal 6 (SIGABRT), code -6 in tid 19193 (FirebaseMLHandl)

This happens when ML Kit can't properly parse the image file—either the JPG is corrupted, the file path/Uri is invalid, or the built-in decoder hits an edge case with your image's encoding.

Second Crash (BMP Format) #

After switching to BMP, the issue shifts to failed model loading, which leads to a segmentation fault:

E/zygote: The String#value field is not present on Android versions >= 6.0
E/native: jni_helper.cc:170 GetContents failed: /data/user/0/com.google.android.gms/app_vision/ocr/data/models/rpn_lstm_engine_tfmini.bincfg
E/native: jni_helper.cc:170 GetContents failed: /data/user/0/com.google.android.gms/app_vision/ocr/data/models/semanticlift_engine_0.2.bincfg
A/libc: Fatal signal 11 (SIGSEGV), code 1, fault addr 0x0 in tid 13092 (ocr_seg_4/13092)

This means the OCR models required by Google Play Services didn't download correctly or can't be accessed, leading to a null pointer crash when the engine tries to use them.

Actionable Fixes #

1. Fix Image Loading (Bypass File Format Issues) #

Instead of using FirebaseVisionImage.fromFilePath(), load the image into a Bitmap first—this standardizes the image data and avoids ML Kit's direct file decoding quirks:

// Replace your existing image initialization code with this
File file = // your image file
Bitmap bitmap = BitmapFactory.decodeFile(file.getAbsolutePath());

// Verify the bitmap loaded successfully
if (bitmap == null) {
    Log.e(TAG, "Failed to load image file - check if the file exists and is not corrupted");
    return;
}

FirebaseVisionImage image = FirebaseVisionImage.fromBitmap(bitmap);

2. Fix OCR Model Loading Failures #

• Update Google Play Services: ML Kit OCR relies on GMS to provide model files. Make sure the device has the latest version of Google Play Services (go to Settings > Apps > Google Play Services > Check for updates).
• Clear GMS Cache: Corrupted model downloads are a common culprit. Go to Settings > Apps > Google Play Services > Storage > Clear Cache, then restart your app—this triggers a fresh download of the OCR models.
• Verify Storage Permissions: If your image is stored in external storage, ensure your app has the READ_EXTERNAL_STORAGE permission (and the user has granted it):

  // Check and request permission if needed
  if (ContextCompat.checkSelfPermission(context, Manifest.permission.READ_EXTERNAL_STORAGE)
          != PackageManager.PERMISSION_GRANTED) {
      ActivityCompat.requestPermissions((Activity) context,
              new String[]{Manifest.permission.READ_EXTERNAL_STORAGE},
              1001); // Use your own request code
  }
  

3. Upgrade to the Latest ML Kit OCR API #

The old VisionTextDetector is deprecated—switch to the newer on-device text recognizer, which is more stable and maintained:

// Initialize the v2 recognizer
FirebaseVisionTextRecognizer recognizer = FirebaseVision.getInstance()
        .getOnDeviceTextRecognizer();

// Process the image (use the Bitmap-based image we created earlier)
Task<FirebaseVisionText> result = recognizer.processImage(image)
        .addOnSuccessListener(firebaseVisionText -> {
            List<String> words = new ArrayList<>();
            for (FirebaseVisionText.Block block : firebaseVisionText.getBlocks()) {
                Log.d(TAG, "Detected Text: " + block.getText());
                for (FirebaseVisionText.Line line : block.getLines()) {
                    for (FirebaseVisionText.Element element : line.getElements()) {
                        String textPart = element.getText();
                        words.add(textPart);
                        Log.d(TAG, "textPart: " + textPart);
                    }
                }
            }
        })
        .addOnFailureListener(e -> {
            Log.e(TAG, "OCR failed", e);
        });

Make sure your app's build.gradle has the latest dependencies:

implementation 'com.google.firebase:firebase-ml-vision:24.1.0'
implementation 'com.google.firebase:firebase-ml-vision-text-recognition-model:20.0.0'

4. Clean Up Resources #

Always close the recognizer/detector when you're done with it to avoid resource leaks:

@Override
protected void onDestroy() {
    super.onDestroy();
    if (recognizer != null) {
        recognizer.close();
    }
}

内容的提问来源于stack exchange，提问作者talha06