Objdetect

Objdetect is a module within OpenCV that provides algorithms and tools for detecting objects in images and video streams. It focuses on cascade classifiers trained with Haar-like features or Local Binary Patterns to identify objects, as well as histogram-based detectors such as the HOG (Histogram of Oriented Gradients) method for pedestrian detection. The module exposes a compact API for loading pre-trained classifiers and applying them to input frames.

Core components include the CascadeClassifier class, which loads XML cascade data (for example, haarcascade_frontalface_default.xml) and performs detection via the detectMultiScale function. This function searches for objects across multiple scales and outputs bounding boxes with optional confidence measures. The HOGDescriptor class and its default people detector offer an alternative, gradient-based approach to detecting pedestrians.

Training and usage: cascades are created via OpenCV's training tools from positive and negative samples and stored as XML files that can be loaded at runtime. Typical usage involves converting the input image to grayscale, applying the cascade, and tuning parameters such as scaleFactor, minNeighbors, and minSize to balance detection rate and false positives. The module is primarily designed for fast, lightweight detection on conventional hardware.

Limitations and context: objdetect methods rely on the quality and diversity of training data and may struggle with occlusion, pose variation, or lighting changes. For modern, robust detection tasks, OpenCV's dnn module supports deep learning detectors that can be more accurate at the cost of computation. objdetect remains widely used for quick, real-time detections on embedded systems and legacy applications.