opencv Using Cascade Classifiers In Java
Syntax
- CascadeClassifier cascade = new CascadeClassifier("cascade.xml"); // Creates a cascade classifier from cascade.xml
- Mat image = Imgcodecs.imread("image.png"); // Converts image.png into a Mat (Matrix) object
- MatOfRect detections = new MatOfRect(); // Creates an empty MatOfRect (Matrix of Rectangles) file, used as output for our detection classes
- detections.toArray(); // Returns an array of Rect objects that can be iterated over
- Imgproc.rectangle(image, new Point(rect.x, rect.y), new Point(rect.x + rect.width, rect.y + rect.height), new Scalar(0, 255, 0)); // Draws an green outlined rectangle from the first Point's x and y locations to the second Point's x and y location onto the Mat object "image". "rect" is a Rect object, usually provided by detections.toArray(). Uses OpenCV's Point class.
- Imgcodecs.imwrite("output.png", image); // Writes the modified Mat object "image" to the "output.png"
- CascadeClassifier.detectMultiScale(image, detections); // Detects any object in the Mat object "image" and outputs the detections in the MatOfRect object "detections"
- CascadeClassifier.detectMultiScale(image, detections, scaleFactor, minNeighbors, flags, minSize, maxSize); // Performs a detection with additional parameters. See details below.
- Imgproc.ellipse(image, center, axes, 0, 0, 360, new Scalar(255, 0, 255), thickness, lineType, 0); // Draws an ellipse onto the image at the point
center. Uses OpenCV's Point class.
Parameters
| Parameter | Details |
|---|---|
| scaleFactor | How much the image size is reduced at each image scale. Default = 1.1 |
| minNeighbors | How many neighbors a candidate rectangle should have before selecting it as an detected object. Default = 4 |
| flags | Legacy flags. In most cases, this should be set to 0. Default = 0 |
| minSize | Minimum size a candidate rectangle can be. This uses OpenCV's Size class. Can be used to decrease detection time and CPU usage, as well as to reduce false positives. |
| maxSize | Maximum size a candidate rectangle can be. This uses OpenCV's Size class. Can be used to decrease detection time and CPU usage, as well as to reduce false positives. |
| axes | Uses OpenCV's Size class. Defines the width and height of the ellipse. |
| thickness | Thickness of the line, in pixels. |
| lineType | Has various parameters. 0 is the solid line, 8 is for a 8-connected line, 4 is for a 4-connected line, and CV_AA is for an antialiased line. Default = 8 |
