Vehicle Object Recognition Using Lidar Ground Height Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle recognition systems face challenges in accurately recognizing objects, especially in conditions like shadows, direct sunlight, strong light sources, and low-light environments, and in congested road situations where the field of view is obstructed, leading to inaccuracies in object shape and location recognition.
Innovation Solution
An object recognition apparatus utilizing a Lidar sensor and processor to determine the type and location of objects, calculate ground height values, and generate a three-dimensional representation using a multiple linear regression model, which helps in accurately recognizing objects even in obstructed views by differentiating between minimum, maximum, and reference ground height values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If image sensors are used for object recognition, then the system can capture visual information, but shadows and light conditions cause misrecognition of objects
Solution Approach 1:
The patent introduces Lidar sensors as an intermediary technology that operates independently of visible light conditions. The Lidar sensor uses laser beams to measure distances and create depth maps, providing complementary information to image sensors. This mediator allows the system to overcome shadow and light interference by relying on active illumination rather than passive light capture.
Solution Approach 2:
The system changes the measurement parameter from two-dimensional image data to three-dimensional spatial information. By using Lidar to obtain depth data and ground height information, the system transforms the problem from recognizing objects in varying light conditions to measuring physical distances and positions, which are immune to lighting variations.
2Measurement precision
If ultrasonic sensors, laser sensors, or Lidar sensors are used to check object existence and distance, then the field of view is obstructed in congested situations, but the shape and location of distant objects cannot be accurately recognized
Solution Approach 1:
The patent transitions from two-dimensional sensor arrays to three-dimensional spatial reconstruction. By using Lidar to generate point clouds and calculate ground height variations, the system creates a vertical dimension (Z-axis) that allows it to see over and around obstacles. The ground trend line analysis in the vertical dimension enables detection of distant objects even when horizontal field of view is blocked.
Solution Approach 2:
The system segments the measurement space into multiple regions: near-field objects, mid-field objects, and distant objects. By analyzing ground height differences and creating separate processing channels for different distance ranges, the system can independently analyze each segment. This allows distant objects to be detected even when near-field objects block the direct line of sight, as the ground trend analysis can infer positions beyond obstacles.
3Measurement precision
If ground height data is processed using simple methods, then processing is fast, but accuracy in congested areas is poor
Solution Approach 1:
The system performs preliminary processing by sorting all detected points according to their horizontal coordinates before analyzing ground height. This pre-organization of data creates an ordered structure that enables efficient sequential processing. By establishing the ground trend line early using sorted points, the system creates a reference framework that speeds up subsequent object detection and height calculation operations.
Solution Approach 2:
The ground trend line calculation serves multiple functions simultaneously: it establishes a reference for measuring object heights, identifies ground variations, and provides a basis for detecting objects at different distances. This self-service approach means one computational structure (the ground trend line) performs multiple measurement tasks, reducing the need for separate processing algorithms and improving overall efficiency.
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 enhances the accuracy of object recognition, particularly in congested areas, reducing the likelihood of collision and improving vehicle safety by providing precise three-dimensional information for autonomous driving and collision warnings.
Implementation Method 1
A vehicle may include a driver assistance device that includes at least one of an ultrasonic sensor, an image sensor, a laser sensor, and a Lidar sensor
Implementation Method 2
recognizing objects such as obstacles in front or outputs a notification about a possibility of a collision with a recognized object
Data Source
AI summary
An embodiment object recognition apparatus includes a Lidar sensor and one or more processors, at least one of the one or more processors being configured to recognize objects based on data received from the Lidar sensor, obtain a minimum value and a maximum value among values of a ground height for each object based on the data, obtain a reference value of the ground height for each object based on the minimum value for each object, and determine an actual value of the ground height for each object based on a difference between the minimum value, the maximum value, and the reference value of the ground height for each object.