Control System of Intelligent Vehicle by Panoramic Vision

摘要

The success of driverless mainly depends on the intelligent visual system which can control car on the road as we human do. A high efficient controlling system based on panoramic vision for intelligent vehicles, supporting 8 cameras at most, is presented in this paper. The visual module is composed of a FPGA and two DSP to carry on the massive high speed parallel processing for panoramic image construction. The system realizes the intelligent vehicle control with real time panoramic image reconstruction and road lane detection. Central controlling module connects all the sub-system by CAN bus, while the visual module is one of the sub-system. During the driving on the road, the lane lines are detected from the panoramic image and tracked on each following frame. Parameters of the deviations are sent to the main control node as soon as the car goes off the course, then the control signal is decoded and sent to the sub-system for further operation such as corresponding steering device or ABS. Bidirectional gated channel on the sub-controlling-system receives real-time signals from sensors of the mechanical positions while the actuator controlling the car to verify the control precision. The correction mechanism responds rapidly to the changing of the lane to keep the car running on the right lane. And the embedded system μC/OS on the central module manages the balance of multitasks for the high speed automatic control of the intelligent vehicle.