Address:

    Room No.- ECE 1110, Electronics and Communication Engineering Discipline, Dr. Syttendranath Academic Building (Academic Building 1), Khulna University, Khulna 9208.

    Email:

    Contact:

    +8801948820119

    Personal Webpage:
    click here

Implementation of Lane-detection Robot Using Fully Connected neural Network

Abstract: Robots are machines capable of sensing its surroundings and carry out a series of pre- assigned tasks with very little or in some cases no direction from a human being. Robots are getting more and more aggravated with the uncovering of each new technology and in recent years, they are being constructed purposively for specific applications to operate in challenging conditions. In the recent era, lane detection in complex road geometry and bad weather conditions and navigating autonomously is a challenging task. Several traditional and state-of-the-art methods has been applied to solve the lane detection problem by many researchers. In this work we have presented a novel lane detection and tracking algorithm as well as a navigating for complex road scenarios based on CNN models and implement it on our robot in real life environment. We have evaluated our model by some measuring parameters (e.g. accuracy curve, loss curve, recall curve, precision curve etc.). And we also carried out some experiments on three different roads. From the accuracy curve it is found, the model is about to 95.25 % accurate and he results of experiments have shown that the lane was detected properly.abstract: Robots are machines capable of sensing its surroundings and carry out a series of pre- assigned tasks with very little or in some cases no direction from a human being. Robots are getting more and more aggravated with the uncovering of each new technology and in recent years, they are being constructed purposively for specific applications to operate in challenging conditions. In the recent era, lane detection in complex road geometry and bad weather conditions and navigating autonomously is a challenging task. Several traditional and state-of-the-art methods has been applied to solve the lane detection problem by many researchers. In this work we have presented a novel lane detection and tracking algorithm as well as a navigating for complex road scenarios based on CNN models and implement it on our robot in real life environment. We have evaluated our model by some measuring parameters (e.g. accuracy curve, loss curve, recall curve, precision curve etc.). And we also carried out some experiments on three different roads. From the accuracy curve it is found, the model is about to 95.25 % accurate and he results of experiments have shown that the lane was detected properly.Abstract: Robots are machines capable of sensing its surroundings and carry out a series of pre- assigned tasks with very little or in some cases no direction from a human being. Robots are getting more and more aggravated with the uncovering of each new technology and in recent years, they are being constructed purposively for specific applications to operate in challenging conditions. In the recent era, lane detection in complex road geometry and bad weather conditions and navigating autonomously is a challenging task. Several traditional and state-of-the-art methods has been applied to solve the lane detection problem by many researchers. In this work we have presented a novel lane detection and tracking algorithm as well as a navigating for complex road scenarios based on CNN models and implement it on our robot in real life environment. We have evaluated our model by some measuring parameters (e.g. accuracy curve, loss curve, recall curve, precision curve etc.). And we also carried out some experiments on three different roads. From the accuracy curve it is found, the model is about to 95.25 % accurate and he results of experiments have shown that the lane was detected properly.

Details
Role Co-Supervisor
Class / Degree Bachelor
Students

Md. Al-Masrur Khan [160902] and  Fatima Tuz Zohora [160919]

Start Date 6th June, 2018
End Date 10th October, 2019