Detection and Tracking of Multiple Moving Objects software package, started in 2016. The main goal of the project is to construct a software package that enables multiple-object recognition for driver-less cars. This project seeks to finally deliver a software package for dynamic object detection and tracking task, based on different sensor modalities. Light Detection And Ranging (LiDAR) sensor, Inertial Measurement Unit (IMU) and onboard Cameras are assumed to construct the sensor modality of a driver-less car. In it’s initial phase, the project was focused on LiDAR-only methodologies to perform dynamic object’s detection and tracking task. Furthermore, both highway and urban environment scenarios are taken into account which are in need of different treatment due to their inherent difference in structure.
The software package developed with the use of Robot Operating System (ROS). Most of libraries are developed in C++ and Python. Package consists of two different modules: Detection and Tracking. In the detection module primary refinement, ground estimation, potent movable object detection and orientation estimation is taking place. The tracking module is then responsible for tracking and data association.
- Primary Refinement
- Ground Estimation (Highway/Rough Areas)
- Potent Movable Object Detection
- Orientation Estimation
- New Track initialization / Old Track prediction
- Data Association
- State Estimation
- Track Management
A Physical-Motivated Three-Dimensional Gaussian Process Based Ground Segmentation Algorithm with Local Characteristic Estimation
Autonomous Land Vehicles (ALV) shall efﬁciently recognize the ground in unknown environments. A Gaussian process based ground segmentation method is proposed in this paper, which is fully developed in a probabilistic framework due to implementation of Bayesian inference. The data is segmented using a radial grid map. Two joint Gaussian processes are introduced to separately model the observation and local characteristics of the data. While, observation process is used to model the ground, the latent process is put on length-scale values to estimate point values of length-scales at each input location. Input locations for this latent process are chosen in a physical motivated procedure to represent an intuition about ground condition. Furthermore, an intuitive guess of length-scale values is represented by assuming of the existence of hypothetical surfaces in the environment that every bunch of data points may be assumed to be resulted from measurements from this surfaces. Bayesian inference is implemented using maximum A posteriori criterion. The log-marginal likelihood function is assumed to be a multitask objective function, to represent a whole-frame unbiased view of the ground at each frame. Simulation results shows the effectiveness of the proposed method even in an uneven, rough scene which outperforms similar Gaussian process-based ground segmentation methods. While adjacent segments do not have similar ground structure in an uneven scene, the proposed method gives an eﬁcient ground estimation based on a whole-frame viewpoint instead of just estimating segment-wise probable ground surfaces.