Week 24 - 31

Achievements made this week:

1. Successful implementation of the Multivariable Segmentation 
2. Representation on the rviz platform of the Segmentation methods proposed, as it is shown on the figure below: 

            Each layer represents the clusters obtained for each Segmentation method (counting upwards):

• Simple segmentation 
• Multivariable Segmentation
• Dietmayer Segmentation
• Adaptative Breakpoint Detector

The next step, is to compare these results with a "groundtruth" and check the errors given by each method when the "threshold" values varies. 

So, a "groundtruth" to make the comparison between the methods is needed.

For that I use a Matlab program made by my Co-Supervisor Jorge Almeida, this program reads the laser scans data form a .txt file, then the data is converted into Cartesian coordinates and creates a structure composed by the coordinates and a label. Then we preform the hand labeling of the data to obtain the target "groundtruth".

The data was collected during a ride to a big roundabout nearby Aveiro. In order to help the hand labeling of the laser data, I use some images took by the car's cameras.  

In 1220 scans (iterations) I identified 17 moving cars and 1 bike.

Meanwhile, I've already started to think in the next phase of the project which is "Feature Extraction Algorithms". So i made a research of some state of art feature extraction algorithms and I've started to implement the algorithm developed by Kai O. Arras in "Using Boosted Features for the Detection of People in 2D Range Data", University of Freiburg.

Week 18 - 23 March

First I've developed a platform for advertising and subscribing a topic from a rosbag which contains sensor_msgs/LaserScan. Them I am applying several Distance-Based Segmentation Algorithms to find out how many objects exist at the time of the scan:

1. Simple Segmentation 
2. Dietmayer Segmentation
3. Adaptative Breakpoint Detector
4. Multivariable Segmentation  - Cosine distance based 

At the moment, none of these algorithms by themselves solves any of the main Tracking Problems such as shape change or occlusions.

After i do the segmentation, i represent the results on a rviz plataform. For each scan, each cluster is represented by a different color and as an id:

   As expected, the results among the three first methods were very similar, because they are all euclidean distance based.

 At the moment i am still trying do improve the implementation Multivariable Segmentation because until now it is showing very poor results.
 

Introduction - Initial Objectives

Greetings,

My name is Daniel Coimbra, i am 22 years old and i am a Mechanical Engineering student from University of Aveiro, Portugal and i am currently working on my master degree project.

The title of my project is "Multiple Target Tracking and Detection on Road Environment ", it is related to the Advanced Preception Systems for the "ATLASCAR", which is an ongoing project of the Laboratory of Automation and Robotics of the Mechanical Engineering Department of the University of Aveiro which the ultimate goal is to build a full autonomous driving car. For more information about the ATLASCAR project, feel free to check the official site: http://atlas.web.ua.pt/atlascar.html

The main  topic of my project relates to the Tracking of moving objects which presents a great importance in several areas:

• Mobile Robotics:
• Calculation of time to collision with surrounding objects, so that dangerous situations may be avoid.
• Path planning for Driving aid systems.
• Security applications:
• Personal access control in buildings.
• Crowd control operations in malls and airport for example.

In short, the main goal of my project is to Develop one or more Multiple Target Tracking Algorithms that should detect the presence of moving agents around the vehicle and then determine the position and velocity, for measuring the distances of the objects from the car i use a Laser Ranger Finder (LRF). The second objective is to extract features and classify them based on the geometric outlines and dynamic behavior. The classification can result in different categories such as: persons, cars, trucks and bikes.

  The main tasks are:

• State of the art in Multiple Target Tracking Algorithms
• Implement different Segmentation Methods
• Simple Segmentation (Euclidean distance and threshold)
• Dietmayer Segmentation
• Adaptative Breakpoint Detector
• Multivariable Segmentation
• K-means
• Implement different feature extraction algorithms
• Geometrical range features extracted from laser-segments and them compose a feature vector to perform the classification.
• Implement a data association algorithm which will associate a measurement to an existing track or a new track.
• Data collection and validation of the Algorithms on the road - real time / on the fly.

I will develop my program using ROS (C++) and RVIZ for data representation,

My first presentation can be found here.