Ultra fast cnn based Hardware Computing Platform Concepts for adas visual Sensors and Evolutionary Mobile Robots

Download 3,22 Mb. Pdf ko'rish
bet	16/81
Sana	16.05.2024
Hajmi	3,22 Mb.
	#238917

1 ... 12 13 14 15 16 17 18 19 ... 81

Bog'liq
Alireza Fasih

Figure 2-1: Comparison of two different ADAS system. (a) Lane departure warning, (b) Licence plate recognition

2.2

State of the art of real-time ADAS platforms
A huge on-going research and several projects are and have been conducted in various
aspects of ADAS technology [39-42]. The sensory system within ADAS does provide all
necessary data and information which is needed to estimate the traffic situation and state
of the vehicle in real-time [43]. For processing images captured by HD cameras [39] people
have been using FPGA and DSP platforms with processing frames with 20~100 FPS.
Examples of systems where this can be observed are [14, 44, 45], lane markings detection
and obstacle detection. One has switched to hardware implementation (i.e. on FPGA or
DSP) in order to reduce the CPU load; see reference [39]. In reference [43], Mr Wada has
been using a normal PC for the development of advanced parking assistance systems.
Further, in reference [46], authors Vitabile
et al
do propose a lane keeping system based on
FPGA that can process up to 39 frames per second; this safety system has been capable of
identifying a dangerous situation and react in real-time. A main drawback of all of
traditional ADAS systems is that they are using different platforms for different tasks.
Thereby a combination of all different tasks does need/consume too much hardware
resources [12, 47].
Figure
2-1: Comparison of two different ADAS system. (a) Lane departure warning, (b)
Licence plate recognition
(b)
Store in DB & Display
Character Recognition
Dilation
Video Capturing
Video Enhancement
Licence Detection
Region of Interest
Dynamic Thresholding
(a)
Warning System
Geometry of Lane
Hough Transfer
Find Edging
Labeling & Filtering
Video Capturing
Video Enhancement
Region of Interest
Dynamic Thresholding
Dilation

21
Figure 2-1 shows the comparison of two different types of ADAS concept. Both systems are
using the same modules of video capturing, region of interest, dynamic thresholding and
dilation. Therefore, without a dynamic architecture we should have to unnecessarily
reserve resources for each of the systems.

Download 3,22 Mb.

1 ... 12 13 14 15 16 17 18 19 ... 81

Download 3,22 Mb.

Pdf ko'rish