Release Notes - Java 2.x
2.2.10 (06/27/2024)
Highlights
-
Improved the performance of the MRZ scenario. The recognition speed is improved by about 70%.
-
Added character normalization modes to normalize the text. This feature will improve the read rate when processing the text in nonstandard fonts like dot matrix.
-
New parameter control and recognition methods are added to fit more usage scenarios.
Changelog
New
- Added a new property
timeout
toDLRRuntimeSettings
class. - Added modes parameter
CharacterNormalizationModes
to normalize the text. The parameter is available under the following classes:LabelRecognizerParameter
TextArea
LineSpecification
Improved
- Reduced the size of MRZ model from 10MB to 2.56MB.
- Improved single-line text confidence. This enables users to implement a result confidence filter to improve the recognition accuracy.
- Improved character segmentation when processing some connected characters. This improves the recognition accuracy.
2.0 (08/26/2021)
Highlights
- Added image processing modes to enhance the usage scenarios like MRZ reading:
- Texture area detection
- Color Conversion
- Binarization
- Grayscale image processing
-
Added dictionary-based correction. The dictionary is minimal and customizable to only cover a scenario-oriented vocabulary.
-
Improved the overall performance by replacing Caffe engine with OpenCV dnn engine
-
Improved the readability of recognizing skewed characters.
-
Improved the MRZ reading performance.
- Added a
core
library. Migrated the core structs/classes from the LabelRecognizer library to thecore
library.
Changelog
New
- Added auto-deskew algorithm to improve the performance on recognizing the skewed characters.
- Added package
dynamsoft-core.jar
. Migrated the Dynamsoft core classes from packagedynamsoft-labelrecognizer.jar
todynamsoft-core.jar
. - Added class
BarcodeResult
for users to interact with Dynamsoft Barcode Reader SDK. - Added
DLRRuntimeSettings
propertydictionaryPath
anddictionaryCorrectionThreshold
for users to further improve the recognizing accuracy by referencing dictionary files. - Added class
DLRDictionaryCorrectionThreshold
. - Added class
DLRFurtherModes
and propertyDLRRuntimeSettings.furtherModes
for users to config more processing modes. - Added enumeration
TextureDetectionMode
and propertyDLRFurtherModes.textureDetectionModes
for users to detect and remove the texture background. - Added enumeration
ColourConversionMode
and propertyDLRFurtherModes.colourConversionModes
for users to convert color images to grayscale images in differenct ways. - Added enumeration
BinarizationMode
and propertyDLRRuntimeSettings.binarizationModes
for users to convert grayscale images to binary images in different ways. - Added enumeration
GrayscaleEnhancementMode
and propertyDLRFurtherModes.grayscaleEnhancementModes
for users to enable grayscale images preprocessing. - Added
characterHConfidence
,characterMConfidence
andcharacterLConfidence
properties inDLRCharacterResult
class so that more alternative results will be available for users.
Improved
- Improved the neural network performance by replacing Caffe engine with OpenCV DNN engine.
Fixed
- Fixed a bug that might cause wrong line number matching when using
LineSpecification.LineNumber
.
API Changes
- Modified the method
initLicense(String)
to staticinitLicense(String)
. - Modified the parameter type of the method
updateReferenceRegionFromBarcodeResults
fromTextResult[]
toBarcodeResult[]
. - Modified the parameters
LabelRecognizerParameter.LetterHeightRange
andTextArea.LetterHeightRange
. The value unit of the parameters are modified from percentage to thousandth. The available range of the value and the default value are updated as well. - Renamed class
LabelRecognition
toLableRecognizer
. - Removed method
initLicenseFromLTS
. - Removed method
LabelRecognizer(string license)
. - Removed method
initLTSConnectionParameters
. - Removed class
DMLTSConnectionParameters
.