LiDAR-Image Fusion Object Recognition for Occluded Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Intelligent vehicles face challenges in accurately recognizing objects due to limitations in existing sensors, such as LiDAR, ultrasonic, and image sensors, particularly in congested areas where the field of vision is obstructed, leading to inaccurate detection of obstacles, road signs, and preceding vehicles.
Innovation Solution
An object recognition device that corrects position and shape information of detected objects using a combination of LiDAR and image sensors, where image information from cameras is used to refine data from LiDAR sensors, especially in interference areas obstructed by adjacent vehicles, to enhance accuracy and visibility.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If LiDAR sensor is used for object recognition, then distance measurement capability is improved, but recognition accuracy deteriorates when objects are obstructed by preceding vehicles
Solution Approach 1:
The patent combines LiDAR sensor data with image sensor data to create a fused recognition result. The LiDAR provides accurate distance information while the image sensor captures visual features, and their integration compensates for the occlusion problem where preceding vehicles block the LiDAR's field of view, thereby maintaining both distance measurement capability and recognition accuracy.
Solution Approach 2:
The image sensor acts as an intermediary to supplement information that the LiDAR cannot obtain due to obstruction. When the LiDAR's field of view is blocked by preceding vehicles, the image sensor captures images of the obstructed areas, and the control unit uses this image information to infer and correct the position and shape of objects that would otherwise be undetectable by LiDAR alone.
2Loss of information
If image sensor is used for object recognition, then visual information is improved, but recognition accuracy deteriorates due to shadows, direct sunlight, and low illuminance
Solution Approach 1:
The patent merges image sensor data with LiDAR sensor data to compensate for the weaknesses of each sensor type. The LiDAR sensor provides reliable distance and shape information that is less affected by lighting conditions, which compensates for the image sensor's vulnerability to shadows, direct sunlight, and low illuminance, thereby maintaining visual information quality and recognition accuracy across varying lighting conditions.
3Reliability
If ultrasonic sensor is used for object recognition, then obstacle detection is improved, but recognition accuracy deteriorates for road signs, ramps, and speed bumps
Solution Approach 1:
The patent combines ultrasonic sensor data with image sensor and LiDAR sensor data to achieve accurate recognition of various objects. The ultrasonic sensor provides reliable obstacle detection capability, while the image sensor captures visual features for identification, and the LiDAR sensor provides precise distance and shape information, together enabling accurate differentiation between obstacles, road signs, ramps, and speed bumps.
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 solution improves the accuracy of object recognition and position detection, even in obstructed areas, enhancing safety and autonomous driving capabilities by accurately identifying surrounding vehicles and obstacles, thus reducing the risk of accidents and improving vehicle performance and user satisfaction.
Implementation Method 1
a distance detection unit that detects a distance to a surrounding object
Implementation Method 2
When the intelligent vehicle recognizes an object using a LiDAR sensor
Implementation Method 3
an image detection unit that detects an image of the surrounding object
Data Source
AI summary
An object recognition device includes: a storage unit that stores image information corresponding to respective shapes of a plurality of objects, and a control unit that groups a plurality of signals detected by a distance detection unit to acquire position information of at least one of the plurality of objects, sets a first object of the plurality of objects based on the acquired position information, sets an interference area based on position information of the first object, acquires position information and image information of a second object of the plurality of objects located in the interference area, and corrects the acquired position information of the second object based on the image information of the second object and the image information stored in the storage unit.