ECE 492 Computer Engineering Design Project Ball Tracking

ECE 492 - Computer Engineering Design Project Ball Tracking and Trajectory Prediction Project Overview Our group created an image processing and image triangulation based projectile detection system with flight trajectory prediction functionalities. The core detection capabilities of our project functions similarly to a dumbed down version of the widely-known Hawk Eye detection system used in professional Tennis matches for line-call disputes. We also implemented a VGA controller to visually demonstrate the prediction ability of our system via a software based dartboard simulation. Video Decoding Tracking Algorithm The DE 2 board provides an ADV 7181 Video Decoder to decode NTSC signals from the camera. It outputs ITU-R 656 coded pixels which we decoded into 30 -bit RGB format to allow for better selectivity and increased tolerance to color fluctuations. To increase the speed of the tracking, we bypass a frame buffer and perform the decoding on the same clock used by the ITU-R 656 decoder. Each DE 2 board can only handle one NTSC input so we used a network of two boards. VGA Controller We implemented external display using a framebuffer based VGA controller. After software generated (320 x 240) frames are written to a frame buffer in SDRAM, a DMA copies the frames into SRAM for VGA video output. Algorithm: Track FOR every line { find orange line IF orange line > threshold d = longest orange line } IF end of frame center = d/2 Figure 2: Ball Tracking Algorithm We created our own ball tracker component in VHDL. Our tracking algorithm consists of scanning every odd line from the ADV 7181 decoder for consecutive matching pixels. We store the longest unbroken sequence of matching pixels in each frame and feed the coordinates of the center of the line to the CPU for processing. The least significant bits of each pixel are affected by miniscule color variations so we only take into account the three most significant bits of each RGB pixel. Figure 1: Project Setup without Display The Setup Our cameras are setup orthogonally with wooden rods and two corner clamps to pre-calibrated view dimensions. Other components include the Altera DE 2 development boards, and an external display connected via VGA, as well as a AA battery package for power. Figure 4: Dartboard Simulation Software System function modes: 1. Tracking mode 2. Prediction Mode Button based interrupts allow the user to navigate through the different modes of our project. VGA frames are created using the software component of our project, using u. C/OSII. X and Y coordinates (from both cameras) of the current location of the tracked object are passed from our tracker component. In tracking mode, a virtual viewing field is simulated via real time movements of a ping pong ball using the cameras, demonstrating functional tracking. In Prediction Mode, mathematical functions calculate and simulate the impact point of a thrown projectile on a virtual dartboard. This is done via triangulation of the data received from the two cameras. Figure 4: Dartboard Simulation Department of Electrical & Computer Engineering