Sensors technologies for intelligent vehicles perception systems: A comparison between vision and 3D-LIDAR

Abstract

3D laser based technology has gained in the past, with the DARPA challenges, a first player place in the arena of sensing technologies for intelligent vehicles. Today, these devices are employed successfully in many new implementations to obtain a 360 degrees view of the area around the vehicle, providing a very dense 3D point cloud that covers an extended range. However, such technology has demonstrated big limitations for being a competitive candidate for off-the-shelf production, in terms of costs and impact on the vehicle design. These requirements are hardly compatible with the constraints dictated by car makers. On the other hand computer vision systems are now being integrated in the cars for many commercial ADAS and are reaching, close enough results in terms of sensing performance to be a viable alternative to the perception benefits offered by the 3D LIDARS. This paper compares pros and cons for these two technologies, using a novel FPGA based stereo vision system implementation, discussing features, and alternatives.