如何替换新版Java OpenCV中已废弃的FeatureDetector函数？

Hey there! I’ve run into this exact issue when migrating OpenCV code from older versions to 3.x and above, so let me break this down clearly for you.

What to Use Instead #

OpenCV 3.x completely redesigned its feature detection API, phasing out the old FeatureDetector abstract class in favor of concrete implementations under the Feature2D interface. Instead of using a factory method with enum constants, you now directly instantiate the specific detector you need via its static create() method.

Here are common examples of how to update your code:

Old (Deprecated) Code #

// This is the outdated approach
FeatureDetector detector = FeatureDetector.create(FeatureDetector.ORB);

New (Recommended) Code #

For standard detectors like ORB, FAST, or BRISK, you can directly create an instance of the detector class:

// Directly instantiate ORB detector
ORB detector = ORB.create();

// If you need a generic reference for flexibility across detectors
Feature2D detector = ORB.create();

For detectors in OpenCV’s extended features module (like SIFT or SURF), use their classes from the org.opencv.xfeatures2d package:

// SIFT detector (requires the opencv_xfeatures2d module)
SIFT detector = SIFT.create();

// SURF detector (also part of the xfeatures2d module)
SURF detector = SURF.create();

The new create() methods also let you configure parameters directly, which is more intuitive than the old set() methods. For example:

// Create ORB with custom parameters (number of features, scale factor, etc.)
ORB detector = ORB.create(500, 1.2f, 8, 31, 0, 2, ORB.HARRIS_SCORE, 31, 20);

Can You Still Use the Deprecated FeatureDetector? #

Technically, yes—most OpenCV versions still support the old FeatureDetector code for backward compatibility. However, it’s strongly not recommended for these reasons:

• The deprecated API will likely be removed in future OpenCV releases, breaking your code when you upgrade.
• The new API is more type-safe and explicit, making your code easier to read and maintain.
• Many new features and parameter improvements are only available in the new Feature2D subclasses.

Key Notes #

• Ensure your OpenCV build includes necessary modules: SIFT/SURF require the opencv_xfeatures2d module (it’s usually included in pre-built Java packages, but double-check if you’re building from source).
• All old detector types have direct replacements in either the main org.opencv.features2d package or the extended xfeatures2d package.

内容的提问来源于stack exchange，提问作者lovecode