LIDAR Object Detection via 2D Birdview Projection and Blurred Template Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle advanced driver assistance systems (ADAS) that utilize LIDAR struggle to distinguish between different types of objects without extensive deep neural networks (DNNs) and substantial processing power, leading to increased costs and decreased efficiency.
Innovation Solution
The system converts 3D LIDAR point cloud data to a 2D birdview projection, blurs both the projection and a template image using an averaging filter, and matches the blurred images for object detection, eliminating the need for DNNs and reducing processing requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep neural networks are used to distinguish different types of objects in LIDAR point clouds, then object detection accuracy is improved, but processing power requirements and costs increase
Solution Approach 1:
The patent transforms the 3D point cloud data into a 2D bird's-eye view representation, changing the dimensional parameters of the data. This transformation simplifies the detection problem by projecting three-dimensional spatial information onto a two-dimensional plane, reducing computational complexity while preserving essential object characteristics for accurate classification
Solution Approach 2:
The patent creates a simplified 2D copy or projection of the original 3D point cloud data. This bird's-eye view representation serves as a computationally efficient surrogate that maintains the essential structural information needed for object detection, allowing accurate classification without processing the full 3D data complexity
2Measurement precision
If deep neural networks with extensive training data are used for object detection, then object classification accuracy is improved, but system complexity and costs increase
Solution Approach 1:
The patent changes the parameter representation from 3D point cloud coordinates to a 2D bird's-eye view projection. This parameter transformation simplifies the data structure and detection algorithm requirements, achieving accurate object classification without the need for complex deep neural network architectures and extensive training datasets
Solution Approach 2:
The patent segments the object detection problem into distinct processing stages: 3D point cloud acquisition, 2D bird's-eye view projection, and template matching. This segmentation allows each stage to be optimized independently, with the 2D projection stage significantly reducing computational requirements compared to end-to-end 3D deep learning approaches
3Measurement precision
If 3D LIDAR point cloud data is processed directly for object detection, then detection accuracy is maintained, but processing time and computational resources increase
Solution Approach 1:
The patent applies a parameter transformation that converts 3D spatial coordinates into a 2D bird's-eye view representation. This parameter change reduces the data dimensionality and computational complexity, significantly improving processing efficiency while maintaining the essential geometric relationships needed for accurate object detection
Solution Approach 2:
The patent transitions from three-dimensional point cloud processing to two-dimensional bird's-eye view analysis. This dimensionality reduction simplifies the computational problem by eliminating the vertical dimension while preserving horizontal spatial relationships, thereby improving processing speed and efficiency without sacrificing detection accuracy
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 performance by softening features in dispersed point cloud data, reduces processing power needs, and simplifies object recognition without requiring complex training data or DNNs, thereby improving vehicle efficiency and reducing costs.
Implementation Method 1
a light detection and ranging (LIDAR) system configured to emit laser light pulses and capture reflected laser light pulses collectively forming three-dimensional (3D) LIDAR point cloud data
Data Source
AI summary
An advanced driver assistance system (ADAS) and method for a vehicle utilize a light detection and ranging (LIDAR) system configured to emit laser light pulses and capture reflected laser light pulses collectively forming three-dimensional (3D) LIDAR point cloud data and a controller configured to receive the 3D LIDAR point cloud data, convert the 3D LIDAR point cloud data to a two-dimensional (2D) birdview projection, obtain a template image for object detection, the template image being representative of a specific object, blur the 2D birdview projection and the template image to obtain a blurred 2D birdview projection and a blurred template image, and detect the specific object by matching a portion of the blurred 2D birdview projection to the blurred template image.