Camera-LiDAR Abnormal Data Filtering for Automotive Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automotive object detection systems are vulnerable to attacks that mislead sensors, leading to incorrect object detection, which can compromise the reliability of Advanced Driver-Assistance Systems (ADAS) and Autonomous Driving (AD) systems.
Innovation Solution
An object detection system for vehicles employs a camera and LIDAR sensor to preprocess and compare image and point cloud data, using similarity models and regression curves to identify and filter abnormal sensor data, ensuring only secure data is fed to the perception engine.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor data from camera and LIDAR is directly fed to the perception engine, then the object detection system operates with simple data processing, but the system becomes vulnerable to attacks that mislead sensors
Solution Approach 1:
The patent applies preliminary action by performing data preprocessing and anomaly detection before the sensor data reaches the perception engine. The system pre-processes image and point cloud data, performs cross-modal consistency checks, and filters out abnormal data in advance, preventing attacked data from misleading the perception engine while maintaining operational reliability
Solution Approach 2:
The patent introduces an intermediary data processing module between the sensors and the perception engine. This intermediary layer performs cross-modal consistency verification by comparing image data with LIDAR point cloud data, identifying anomalies through regression curve analysis, and filtering suspicious data before it reaches the perception engine, thus acting as a mediator that enhances reliability without significantly increasing overall system complexity
2Measurement precision
If the system processes and compares image and point cloud data to detect abnormalities, then the detection accuracy improves, but the processing time increases
Solution Approach 1:
The patent applies segmentation by dividing the data processing into distinct modular stages: image preprocessing, point cloud preprocessing, cross-modal consistency verification, regression curve-based anomaly detection, and filtering. Each stage processes specific aspects of the data independently, enabling parallel processing and optimizing the balance between detection accuracy and processing time
Solution Approach 2:
The patent implements partial action by selectively applying comprehensive cross-modal verification only to regions or data points that show initial signs of inconsistency or are identified as high-priority areas. Not all data points undergo the full anomaly detection pipeline, reducing overall processing time while maintaining high detection accuracy for suspicious data
Data Source
AI summary
An object detection system of a vehicle includes a camera and a LIDAR sensor. The camera and the LIDAR sensor sense an environment to generate an image and a point cloud that depict the environment. The image and point cloud are preprocessed to facilitate comparison between the image and the point cloud. Similarity between the image and the point cloud in depicting the environment is determined to detect abnormal sensor data. Abnormal sensor data is further detected based on directional pattern strengths of edges of the image and expanded points of the point cloud. Detected abnormal sensor data in the image and point cloud are filtered to generate a secure image and a secure point cloud, which are provided to a perception engine to detect objects or other features in the environment.