我正在尝试整合anyline ocr sdk以从图像中提取url链接。 从他们的documentation我了解到，通过修改iban扫描仪示例，我可以获得相同的网址结果。
以下是扫描视图的配置文件
{
"captureResolution":"1080",
"cutout": {
"style": "rect",
"maxWidthPercent": "80%",
"maxHeightPercent": "80%",
"alignment": "top_half",
"width": 900,
"ratioFromSize": {
"width": 10,
"height": 1
},
"strokeWidth": 2,
"cornerRadius": 10,
"strokeColor": "FFFFFF",
"outerColor": "000000",
"outerAlpha": 0.3,
"feedbackStrokeColor": "0099FF"
},
"flash": {
"mode": "manual",
"alignment": "bottom_right"
},
"beepOnResult": true,
"vibrateOnResult": true,
"blinkAnimationOnResult": true,
"cancelOnResult": true,
"visualFeedback": {
"style": "contour_point",
"strokeColor": "0099FF",
"strokeWidth": 2,
"fillColor": "110099FF"
}
}
ScanURLActivity.java代码
public class ScanURLActivity extends AppCompatActivity {
private static final String TAG = ScanURLActivity.class.getSimpleName();
private AnylineOcrScanView scanView;
private URLResultView urlResultView;
@Override
protected void onCreate(Bundle savedInstanceState) {
super.onCreate(savedInstanceState);
//Set the flag to keep the screen on (otherwise the screen may go dark during scanning)
getWindow().addFlags(WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON);
setContentView(R.layout.activity_anyline_ocr);
addURLResultView();
String license = getString(R.string.anyline_license_key);
// Copies given traineddata-file to a place where the core can access it.
// This MUST be called for every traineddata file that is used (before startScanning() is called).
// The file must be located directly in the assets directory (or in tessdata/ but no other folders are allowed)
scanView = new AnylineOcrScanView(getApplicationContext(),null);
scanView.copyTrainedData("tessdata/eng_no_dict.traineddata", "d142032d86da1be4dbe22dce2eec18d7");
scanView.copyTrainedData("tessdata/deu.traineddata", "2d5190b9b62e28fa6d17b728ca195776");
//Configure the OCR for URLs
AnylineOcrConfig anylineOcrConfig = new AnylineOcrConfig();
// use the line mode (line length and font may vary)
anylineOcrConfig.setScanMode(AnylineOcrConfig.ScanMode.LINE);
// set the languages used for OCR
anylineOcrConfig.setTesseractLanguages("eng_no_dict", "deu");
// allow only capital letters and numbers
anylineOcrConfig.setCharWhitelist("ABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890abcdefghijklmnopqrstuvwxyz.\\:");
// set the height range the text can have
anylineOcrConfig.setMinCharHeight(20);
anylineOcrConfig.setMaxCharHeight(60);
// The minimum confidence required to return a result, a value between 0 and 100.
// (higher confidence means less likely to get a wrong result, but may be slower to get a result)
anylineOcrConfig.setMinConfidence(65);
// a simple regex for a basic validation of the URL, results that don't match this, will not be returned
// (full validation is more complex, as different countries have different formats)
anylineOcrConfig.setValidationRegex("^(?:http(s)?:\\/\\/)?[\\w.-]+(?:\\.[\\w\\.-]+)+[\\w\\-\\._~:/?#[\\]@!\\$&'\\(\\)\\*\\+,;=.]+$");
// removes small contours (helpful in this case as no letters with small artifacts are allowed, like iöäü)
anylineOcrConfig.setRemoveSmallContours(true);
// removes whitespaces from the result
// (also causes faster processing, because optimizations can be made if whitespaces are not relevant)
anylineOcrConfig.setRemoveWhitespaces(true);
// Experimental parameter to set the minimum sharpness (value between 0-100; 0 to turn sharpness detection off)
// The goal of the minimum sharpness is to avoid a time consuming ocr step,
// if the image is blurry and good results are therefor not likely.
anylineOcrConfig.setMinSharpness(66);
// set the ocr config
scanView.setAnylineOcrConfig(anylineOcrConfig);
// set an individual focus configuration for this example
FocusConfig focusConfig = new FocusConfig.Builder()
.setDefaultMode(Camera.Parameters.FOCUS_MODE_AUTO) // set default focus mode to be auto focus
.setAutoFocusInterval(8000) // set an interval of 8 seconds for auto focus
.setEnableFocusOnTouch(true) // enable focus on touch functionality
.setEnablePhaseAutoFocus(true) // enable phase focus for faster focusing on new devices
.setEnableFocusAreas(true) // enable focus areas to coincide with the cutout
.build();
// set the focus config
scanView.setFocusConfig(focusConfig);
// set the highest possible preview fps range
scanView.setUseMaxFpsRange(true);
// set sports scene mode to try and bump up the fps count even more
scanView.setSceneMode(Camera.Parameters.SCENE_MODE_SPORTS);
// initialize with the license and a listener
scanView.initAnyline(license, new AnylineOcrListener() {
@Override
public void onReport(String identifier, Object value) {
// Called with interesting values, that arise during processing.
// Some possibly reported values:
//
// $brightness - the brightness of the center region of the cutout as a float value
// $confidence - the confidence, an Integer value between 0 and 100
// $thresholdedImage - the current image transformed into black and white
// $sharpness - the detected sharpness value (only reported if minSharpness > 0)
}
@Override
public boolean onTextOutlineDetected(List<PointF> list) {
// Called when the outline of a possible text is detected.
// If false is returned, the outline is drawn automatically.
return false;
}
@Override
public void onResult(AnylineOcrResult result) {
// Called when a valid result is found (minimum confidence is exceeded and validation with regex was ok)
urlResultView.setResult(result.getText());
urlResultView.setVisibility(View.VISIBLE);
}
@Override
public void onAbortRun(AnylineOcrError code, String message) {
// Is called when no result was found for the current image.
// E.g. if no text was found or the result is not valid.
}
});
// disable the reporting if set to off in preferences
if (!PreferenceManager.getDefaultSharedPreferences(this).getBoolean(
SettingsFragment.KEY_PREF_REPORTING_ON, true)) {
// The reporting of results - including the photo of a scanned meter -
// helps us in improving our product, and the customer experience.
// However, if you wish to turn off this reporting feature, you can do it like this:
scanView.setReportingEnabled(false);
}
urlResultView.setOnClickListener(new View.OnClickListener() {
@Override
public void onClick(View v) {
urlResultView.setVisibility(View.INVISIBLE);
scanView.startScanning();
}
});
}
private void addURLResultView() {
RelativeLayout mainLayout = (RelativeLayout) findViewById(R.id.main_layout);
RelativeLayout.LayoutParams params = new RelativeLayout.LayoutParams(
ViewGroup.LayoutParams.WRAP_CONTENT, ViewGroup.LayoutParams.WRAP_CONTENT);
params.addRule(RelativeLayout.CENTER_HORIZONTAL, RelativeLayout.TRUE);
params.addRule(RelativeLayout.CENTER_VERTICAL, RelativeLayout.TRUE);
urlResultView = new URLResultView(this);
urlResultView.setVisibility(View.INVISIBLE);
mainLayout.addView(urlResultView, params);
}
@Override
protected void onResume() {
super.onResume();
scanView.startScanning();
}
@Override
protected void onPause() {
super.onPause();
scanView.cancelScanning();
scanView.releaseCameraInBackground();
}
@Override
public void onBackPressed() {
if (urlResultView.getVisibility() == View.VISIBLE) {
urlResultView.setVisibility(View.INVISIBLE);
scanView.startScanning();
} else {
super.onBackPressed();
}
}
}
尽管置信度设置为65，AnylineOCRlistener仍无法检测到任何ocr结果。
URLResultView.java类的代码
public class URLResultView extends RelativeLayout {
private TextView resultText;
public URLResultView(Context context) {
super(context);
init();
}
public URLResultView(Context context, AttributeSet attrs) {
super(context, attrs);
init();
}
public URLResultView(Context context, AttributeSet attrs, int defStyleAttr) {
super(context, attrs, defStyleAttr);
init();
}
private void init() {
setPadding(DimensUtil.getPixFromDp(getContext(), 4), DimensUtil.getPixFromDp(getContext(), 16),
DimensUtil.getPixFromDp(getContext(), 4), DimensUtil.getPixFromDp(getContext(), 16));
//setBackgroundResource(R.drawable.);
inflate(getContext(), R.layout.url_result, this);
resultText = (TextView) findViewById(R.id.text_result);
}
public void setResult(String result) {
resultText.setText(result.trim());
}
}
有人可以帮我解决这个问题，因为除了他们的文档和sdk样本之外我找不到任何其他资源/教程。
答案 0 :(得分：-1)
实际上您遇到了两个问题：
\添加了反斜杠charWhitelist，而不是斜杠/ removeSmallContours设置为false removeSmallContours会从小于minCharHeight的行中删除所有内容。因此，在您的情况下，它会删除:和.，因为它们对于SDK来说太小了。
更改这两个设置后，扫描工作正常，minConfidence为85.