Customise Your Training
TechSource Systems is MathWorks Authorised Reseller and Training Partner
Automated Driving with MATLAB
Learn how to develop and verify automated driving perception algorithms in MATLAB
Course
Highlights
An automated driving system primarily consists of four main components: perception, localization, planning and control. Automated Driving ToolboxTM focuses on perception which consists of all the low level sensor processing such as camera, radar, and lidars to produce high level outputs such as object detections, lane detections, etc. Some of common problems that is tackled in automated driving such as how to visualize vehicle data, how to detect lanes and vehicles, how to fuse detections from multiple sensors.
This two-day course provides hands-on experience with developing and verifying automated driving perception algorithms. Examples and exercises demonstrate the use of appropriate MATLAB® and Automated Driving Toolbox™ functionality. Topics include:
- Labeling of ground truth data
- Visualizing sensor data
- Detecting lanes and vehicles
- Processing lidar point clouds
- Tracking and sensor fusion
- Generating driving scenarios and modeling sensors
Who Should
Attend
Engineers who develop automated driving systems.
Course
Prerequisites
MATLAB Fundamentals or equivalent experience using MATLAB. Image Processing with MATLAB, Computer Vision with MATLAB and basic knowledge of image processing and computer vision concepts. Deep Learning with MATLAB is recommended.
Course
Benefits
Upon the completion of the course, the participants will be able to:
- automate labeling and creating ground truth
- visualize sensor detection in bird’s eye view and camera perspectives
- detect lane boundaries and compute boundary models
- detect vehicles using pretrained object detectors
- segment, align and combine lidar point clouds
- fuse and track detection from various sensors
- specify and simulate driving scenarios
- create statistical models for generating synthetic camera and radar detections
Partners
Upcoming Program
- Please keep me posted on the next schedule
- Please contact me to arrange customized/ in-house training
Course Outline
Labeling of Ground Truth Data
Objective: Label ground truth data in a video or sequence of images interactively. Automate the labeling with detection and tracking algorithms.
- Overview of the Ground Truth Labeler app
- Label regions of interest (ROIs) and scenes
- Automate labeling
- View and export ground truth results
Visualizing Sensor Data
Objective: Visualize camera frames, radar, and lidar detections. Use appropriate coordinate systems to transform image coordinates to vehicle coordinates and vice versa.
- Create a bird’s-eye plot
- Plot sensor coverage areas
- Visualize detections and lanes
- Convert from vehicle to image coordinates
- Annotate video with detections and lane boundaries
Detecting Lanes and Vehicles
Objective: Segment and model parabolic lane boundaries. Use pretrained object detectors to detect vehicles.
- Perform a bird’s-eye view transform
- Detect lane features
- Compute lane model
- Validate lane detection with ground truth
- Detect vehicles with pretrained object detectors
Processing Lidar Point Clouds
Objective: Work with lidar data stored as 3-D point clouds. Import, visualize, and process point clouds by segmenting them into clusters. Register point clouds to align and build an accumulated point cloud map.
- Import and visualize point clouds
- Preprocess point clouds
- Segment objects from lidar sensor data
- Build a map from lidar sensor data
Fusing Sensor Detections and Tracking
Objective: Create a multi-object tracker to fuse information from multiple sensors such as camera, radar and lidar.
- Track multiple objects
- Preprocess detections
- Utilize Kalman filters
- Manage multiple tracks
- Track with multi-object tracker
Tracking Extended Objects
Objective: Create a probability hypothesis density tracker to track extended objects and estimate their spatial extent.
- Define sensor configurations
- Track extended objects
- Estimate spatial extent
Generating Driving Scenarios and Modeling Sensors
Objective: Create driving scenarios and synthetic radar and camera sensor detections interactively to test automated driving perception algorithms.
- Overview of the Driving Scenario Designer app
- Create scenarios with roads, actors, and sensors
- Simulate and visualize scenarios
- Generate detections and export scenarios
- Test algorithms with scenarios
