Lidar Split-Return ML Classification Against Particulate False Positives
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Lidar systems face false positive detections due to reflections from particulate matter such as fog, smoke, and steam, and split returns caused by edges of surfaces, which can lead to safety risks and operational inaccuracies.
Innovation Solution
A machine-learned (ML) model classifies split returns as associated with particulate matter or true positive surfaces using edge detection and image classification, and adjusts detection thresholds based on return density and image analysis to suppress false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the lidar system detects all returned light signals, then the detection sensitivity is improved, but false positive detections increase due to particulate matter reflections
Solution Approach 1:
The patent segments the detection process into multiple stages: initial signal detection, split return identification, ML classification, and threshold adjustment. By dividing the detection workflow into distinct phases with different criteria, the system maintains high sensitivity while filtering out false positives from particulate matter.
Solution Approach 2:
The machine learning model acts as an intermediary between raw signal detection and final object identification. This intermediary layer classifies split returns by analyzing patterns and contexts, determining whether they represent true surfaces or particulate matter reflections before confirming detections.
2Adaptability or versatility
If the detection threshold is lowered to detect occluded objects, then the detection coverage is improved, but false positives from particulate matter increase
Solution Approach 1:
The detection threshold is made dynamic rather than static. The system adjusts thresholds based on return density, split return patterns, and ML classification results. This allows the threshold to be lowered in contexts suggesting true objects while remaining higher when particulate matter is detected, achieving both coverage and reliability.
Solution Approach 2:
The system changes multiple detection parameters simultaneously: threshold values, return density requirements, and classification confidence levels. By coordinating changes across these parameters, the system adapts to different environmental conditions while maintaining discrimination between true objects and particulate matter.
3Reliability
If the lidar system uses split return analysis to filter false positives, then the false positive rate is reduced, but the device complexity increases
Solution Approach 1:
The system uses the split return signals themselves as the filtering mechanism rather than requiring external validation sensors. By analyzing the temporal and spatial characteristics of split returns, the ML model self-distinguishes between true surfaces and particulate matter using only the existing lidar data, avoiding additional hardware complexity.
4Measurement precision
If machine learning classification is applied to all returns, then the detection accuracy is improved, but the processing time increases
Solution Approach 1:
The ML classification is applied selectively rather than universally. The system identifies split returns and applies classification primarily to these ambiguous cases, while using simpler rules for clear detections. This partial application of ML maintains accuracy for difficult cases while minimizing processing overhead for straightforward detections.
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
Improves lidar detection accuracy by reducing false positives and ensuring accurate identification of solid surfaces, enhancing safety by enabling vehicles to detect occluded objects.
Implementation Method 1
A lidar system has a light emitter and a light sensor. The light emitter may comprise a laser that directs light into an environment. When the emitted light is incident on a surface, a portion of the light is reflected and received by the light sensor
Implementation Method 2
the light sensor, which converts light intensity to a corresponding electrical signal
Implementation Method 3
the system may measure the propagation time of a light signal as it travels from the laser emitter, to the surface, and back to the light sensor. A distance is then calculated based on the flight time and the known speed of light
Implementation Method 4
fine particulate matter may also reflect light. Problematically, fog, smoke, exhaust, steam, and other such vapors may reflect light emitted by a lidar system
Data Source
AI summary
A machine-learned (ML) model for detecting that depth data (e.g., lidar data, radar data) comprises a split beam as opposed to a dual return caused by particulate matter, such as dust, steam, smoke, rain, etc. may classify split beam returns as forming a hollow shape or as forming a solid shape. The ML model may classify split returns forming hollow shapes as being associated with multiple solid surfaces, whereas split returns forming solid shapes may be classified as being associated with particulate matter. Determining a cluster, centroid, density, an image detection, and/or the like may additionally or alternatively be used as input to the ML model and/or to weight outputs of the ml model.