ObjectDetector

Object detection and tracking for pick-and-place robotics operations — used in the paper "Object Detection and Tracking for Pick and Place Robotics Operations".

File Descriptions

• object_detection/sobel_detector.py — ROS2 node (JengaDetectorNode) that subscribes to a RealSense D405 camera, detects Jenga blocks via HSV color segmentation and contour analysis, computes 3D positions using aligned depth data, and publishes annotated images and PoseArray messages for robot arm control.

• object_detection/jenga_moveit_node.py — ROS2 node (JengaCombinedNode) that controls an Interbotix PX150 arm via MoveIt2. Implements a state machine: D405 tracking → ESP32-CAM fine waist correction → grasp sequence (open gripper, hover, descend, close, lift). Integrates live waist nudges from the ESP32-CAM offset during the grasp.

• object_detection/esp_image_detector.py — ROS2 node (SimpleJengaDetector) that processes ESP32-CAM images, detects Jenga blocks with HSV filtering and contour scoring, and publishes a normalised horizontal offset (Float32) used by the arm node for waist joint correction.

• object_detection/cam_driver.py — ROS2 node (Esp32CamDriver) that drives an ESP32-CAM module: fetches an MJPEG HTTP stream, decodes JPEG frames, and publishes them as sensor_msgs/Image topics. Supports start/stop streaming services and rosbag recording/replay.