Birds-Eye-View Lidar Map Object Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated systems face challenges in real-time processing of sensor data to effectively detect and differentiate between dynamic and static objects in their environment, which is crucial for navigation and control.
Innovation Solution
The implementation of a system that uses a lidar sensor to generate a birds-eye-view map, which is then processed by a convolutional neural network to classify objects as dynamic or static, enabling an object tracker to focus on moving objects and inform motion planning, thereby improving navigation and control.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automated systems process sensor data in real-time to detect and differentiate between dynamic and static objects, then navigation and control performance is improved, but processing complexity and computational load increase
Solution Approach 1:
The system segments the processing task by separating dynamic object detection from static object detection. It first identifies static objects (barriers, curbs, sidewalks) and excludes them from dynamic object tracking, thereby simplifying the overall processing complexity while maintaining detection accuracy.
Solution Approach 2:
The system extracts and removes static objects from the sensor data processing pipeline after initial identification. By taking out static objects that do not require tracking, the system reduces computational load on the dynamic object tracker while preserving navigation safety.
2Productivity
If the system classifies all detected objects to distinguish dynamic from static, then tracking efficiency is improved, but processing time increases
Solution Approach 1:
The system performs preliminary classification of objects as static or dynamic before initiating tracking. By pre-identifying static objects and excluding them from the tracking pipeline, the system reduces the number of objects requiring continuous tracking, thereby improving tracking efficiency without significantly increasing processing time.
3Loss of information
If the system processes all sensor data without classification, then no information is lost, but computational resources are wasted on static objects
Solution Approach 1:
The system extracts static objects from the complete sensor data set after initial detection and classification. By removing static objects from further processing, the system prevents waste of computational energy on objects that do not require tracking, while maintaining complete information about the environment for navigation purposes.
Solution Approach 2:
The system applies partial processing to static objects (detection and classification only) versus full processing to dynamic objects (detection, classification, and continuous tracking). This selective processing approach reduces overall computational energy consumption while maintaining necessary information completeness.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enhances the accuracy and efficiency of object recognition and tracking, leading to improved performance in automated vehicle control systems by effectively distinguishing and managing dynamic and static objects in real-time.
Implementation Method 1
a lidar sensor configured to detect objects in a space near the system
Implementation Method 2
one or more objects that are reflected in the lidar data
Data Source
AI summary
Systems and methods for barrier detection using birds-eye-view maps of lidar data. Some implementations may include receiving lidar data from a lidar sensor; determining, based on the lidar data, a birds-eye-view map of one or more objects that are reflected in the lidar data; and inputting the birds-eye-view map to a machine learning system to obtain a classified map wherein the one or more objects are labeled with respective classifications from a set of classes including a dynamic class and static class.