International Journal of Automation and Smart Technology 1 Volume 13 | Issue 1 | 2366 ORIGINAL ARTICLE Boarding and Acceleration on FPGA of an Obstacle Detection Module over Stereo Image Mohamed Salah Salhi*, Manel Salhi, and Hamid Amiri University of Tunis El Manar, National Engineering School of Tunis-Tunisia Research Laboratory of Signal Image and Information Technology LR-SITI. *Corresponding author: medsalah.salhi@enit.utm.tn Accepted: 1 st May 2023 OPEN ACCESS Abstract: This paper presents an accelerated algorithm realized on field programming gate array FPGA to mimic the human vision system. The FPGA is chosen as support owing to their real time propriety RTP. The proposed approach aims to detect any natural image and obtain a "cyclopean image" that provides information on the objects depth with a better details resolution. This strategy uses the common advantages of hybrid stereo imaging models in processing two images of the same scene captured from different viewing angles. Then, it merges the two views to produce a final image containing the depth information of different objects. The adopted algorithm is applied to obstacles detection in which execution time is experimented and shown in results below. Keywords: Accelerated algorithm; Stereo imaging models; Cyclopean image; Real time propriety RTP. Introduction The stereo image history is linked to the human vision way, which exhibits many interesting properties. One of them is the ability to feel the depth of the scenes seen. The process behind this ability is called “stereopsis”. In fact, the brain processes the images from the left and right eyes and executes an image called a "cyclopean image". The cyclopean image then provides information on the objects depth with a better details resolution. In order to mimic the human vision system, stereo imaging models have been created. The stereo vision system processes two images of the same scene captured from different viewing angles. Then, it merges the two views to produce a final image containing the depth information of different objects. Stereo imaging has found its application in many fields such as engineering, architecture, science, education and the military. We can find systems based on stereo imagery in 3D computer graphics, simulators, training systems and robotics. The primitives and dimensions established by a stereo image have been widely applied in home automation, robotic and other security measures. A growing field of application is the autonomous driving of robots through guidance systems. The stereo imaging sensors used in this application type are the most often mounted in a "fish-eyes» topology. They, therefore, have good characteristics such as the wide detection angle: 3/4 of a surrounding circle, with vertical and horizontal resolution, the ability to detect obstacles in a short time and the simultaneous detection of different obstacles with information about their positions and shapes. In this paper, we try to take advantage of stereo imagery to develop a ground obstacle detection module. This module will allow a robotic system to sense its environment to make better decisions while determining its orientation. Indeed, ground is considered a reference when detecting obstacles and determining their real position. Many methods have been used to tackle the ground plane estimation, such as Least Squares LS, Eigen Value EV, and Expectation Maximization Methods EMM, etc [1]. Ground plane estimation is about distinguishing the ground pixels in an image. For an application, like pedestrian detection, ground estimation is important. This paper is presented under different sections