Occupancy Grid Fusion for Radar and Camera Obstacle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing obstacle detection systems using vision-imaging and radar systems often face uncertainty when combining data from different sensors, as significant discrepancies in detection and tracking results from one device versus another lead to unclear data reliability for identifying obstacles.
Innovation Solution
The system fuses radar and imaging data at the sensor level by partitioning the field of view into an occupancy grid, extracting features from each cell using both sensor types, and employing a primary classifier to determine obstacles, thereby increasing accuracy in obstacle identification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If detection and tracking are performed independently by each sensor then results are combined, then device complexity is reduced, but measurement precision deteriorates due to uncertainty in data reliability when detection results differ significantly between sensors
Solution Approach 1:
The patent segments the fusion process into distinct stages: sensor-level feature extraction from occupancy grids, intermediate feature combination, and final detection/tracking. This segmentation allows each sensor's data to be processed independently first, then combined at the feature level where discrepancies can be resolved more effectively, maintaining lower complexity while improving precision.
Solution Approach 2:
The patent introduces an intermediary feature extraction stage that processes sensor data before final fusion. By extracting features from occupancy grids and combining them at the feature level rather than directly fusing raw sensor data or final detection results, the system creates an intermediate representation that reduces uncertainty and improves obstacle identification accuracy.
2Measurement precision
If sensor data is fused at the sensor level, then measurement precision improves through richer information content, but device complexity increases due to the fusion process requirements
Solution Approach 1:
The fusion process is segmented into manageable stages: occupancy grid generation from raw sensor data, feature extraction from grids, feature combination, and detection/tracking. This segmentation reduces the overall complexity by breaking down the complex fusion task into simpler, more manageable processing steps while still achieving sensor-level fusion benefits.
Solution Approach 2:
The patent extracts essential features from sensor data through occupancy grids before fusion, rather than fusing all raw sensor data directly. This extraction process removes redundant and irrelevant information, reducing the complexity of the fusion process while preserving the critical information needed for accurate obstacle identification.
3Reliability
If detection results from different sensors are combined when they differ significantly, then completeness of obstacle detection is improved, but reliability deteriorates due to uncertainty in which data is more correct
Solution Approach 1:
The occupancy grid serves as an intermediary structure that standardizes sensor data from different sources into a common representation. By converting radar and imaging data into occupancy grids with consistent spatial resolution and coordinate systems, the system reduces uncertainty when combining detection results, as all data is expressed in the same framework.
Solution Approach 2:
The patent transforms sensor data into occupancy grids with specific parameter specifications (spatial resolution, coordinate system, probability values). This parameter standardization allows for more reliable combination of detection results from different sensors, as the data is expressed in consistent terms that facilitate direct comparison and fusion.
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 of obstacle detection by combining sensor data at the sensor level, improving the confidence in identifying obstacles and enabling better integration of radar and imaging data for enhanced vehicle safety applications.
Implementation Method 1
Radar systems utilize radio waves to determine the range, altitude, direction, or speed of objects. A transmitter transmits pulses of radio waves which bounce off any object in their path. The pulses reflected off the object returns a small part of the radio wave's energy to a receiver
Data Source
AI summary
A vehicle obstacle detection system includes an imaging system for capturing objects in a field of view and a radar device for sensing objects in a substantially same field of view. The substantially same field of view is partitioned into an occupancy grid having a plurality of observation cells. A fusion module receives radar data from the radar device and imaging data from the imaging system. The fusion module projects the occupancy grid and associated radar data onto the captured image. The fusion module extracts features from each corresponding cell using sensor data from the radar device and imaging data from the imaging system. A primary classifier determines whether an extracted feature extracted from a respective observation cell is an obstacle.