Hybrid-View LIDAR Object Detection for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately detecting and classifying objects proximate to them, particularly in distinguishing smaller objects from larger ones when they are close, which affects motion planning and safety.
Innovation Solution
A hybrid-view LIDAR-based object detection system that processes both range-view and top-view representations of LIDAR data using a machine-learned detector model to detect objects and generate bounding shapes, improving processing speed and accuracy by analyzing denser data sets and correlating representations for precise object segmentation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single-view LIDAR detection system is used, then the device complexity is reduced, but the object detection accuracy deteriorates when distinguishing smaller objects from larger ones in close proximity
Solution Approach 1:
The patent transforms LIDAR data from a single-range view into multiple representations including top-view, side-view, and front-view projections. This dimensional transformation allows objects to be viewed from different perspectives, enabling better distinction between small and large objects in close proximity by exploiting their different spatial signatures across multiple views.
Solution Approach 2:
The patent segments the LIDAR point cloud data into multiple view representations (top-view, side-view, front-view) and processes each view separately through neural network detectors. This segmentation allows the system to analyze objects from different geometric perspectives, improving detection accuracy without requiring a single complex monolithic detector.
2Measurement precision
If multiple LIDAR representations are processed, then object detection accuracy is improved, but the processing time increases
Solution Approach 1:
The patent divides the processing task into parallel segments where multiple neural network detectors simultaneously process different view representations (top-view, side-view, front-view). This parallel segmentation of the detection pipeline enables the system to leverage multiple views for improved accuracy while maintaining real-time processing throughput.
3Measurement precision
If denser LIDAR data sets are analyzed, then object detection accuracy is improved, but the computational load increases
Solution Approach 1:
The patent extracts and processes only the relevant features from dense LIDAR point cloud data through multiple view projections and neural network detectors. By transforming the full 3D point cloud into 2D projections and using learned feature extraction, the system reduces computational load while retaining the essential information needed for accurate object detection.
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 object detection accuracy, reduces latency, and improves passenger safety by enabling more precise motion planning and vehicle control, effectively distinguishing smaller objects from larger ones in close proximity.
Implementation Method 1
LIDAR data from one or more LIDAR systems configured to transmit ranging signals
Data Source
AI summary
Systems and methods for detecting and classifying objects proximate to an autonomous vehicle can include a sensor system and a vehicle computing system. The sensor system includes at least one LIDAR system configured to transmit ranging signals relative to the autonomous vehicle and to generate LIDAR data. The vehicle computing system receives the LIDAR data from the sensor system. The vehicle computing system also determines at least a range-view representation of the LIDAR data and a top-view representation of the LIDAR data, wherein the range-view representation contains a fewer number of total data points than the top-view representation. The vehicle computing system further detects objects of interest in the range-view representation of the LIDAR data and generates a bounding shape for each of the detected objects of interest in the top-view representation of the LIDAR data.