3D Camera Row Navigation Under Degraded GNSS Signals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic steering systems for vehicles face challenges in precise navigation due to degraded GPS signals from obstructions like trees and buildings, and imaging systems can drift in accuracy when identifying row boundaries and vehicle position, especially with discontinuities in plant growth.
Innovation Solution
A 3-D camera system integrated with visual odometry (VO) and global navigation satellite system (GNSS) data to accurately steer vehicles by generating 3-D point cloud maps, identifying row centers, and performing simultaneous localization and mapping (SLAM) to navigate through fields even without high-precision GPS, using sensor fusion to detect obstructions and end-of-row locations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If wheel odometry and inertial navigation systems are used to compensate for degraded GPS signals, then vehicle position can be maintained, but the system complexity increases and drift accumulates over time
Solution Approach 1:
The patent combines multiple sensing modalities (GNSS receiver, 3-D camera, inertial measurement unit) into an integrated sensor fusion system that processes data from all sources simultaneously. The controller fuses GNSS position data with visual odometry from the 3-D camera and inertial data to maintain accurate vehicle position even when GPS signals are degraded, while avoiding the drift accumulation problem of standalone inertial systems.
Solution Approach 2:
The 3-D camera system acts as an intermediary between degraded GNSS signals and the vehicle control system. By capturing images of row structures and performing visual odometry, the camera provides intermediate position and orientation data that bridges gaps in GNSS coverage, allowing the controller to maintain navigation accuracy without relying solely on complex inertial navigation systems.
2Measurement precision
If imaging systems are used to identify row boundaries and vehicle position, then navigation can be achieved, but accuracy degrades when plants form gaps or extend over adjacent rows
Solution Approach 1:
The patent segments the image processing into multiple independent analysis streams: one for detecting row structures (lines representing rows), another for identifying plant patterns, and a third for determining vehicle position. The controller integrates results from all streams, allowing the system to maintain row boundary detection accuracy even when individual plant patterns are discontinuous or irregular.
Solution Approach 2:
The system transitions from 2-D image analysis to 3-D spatial reasoning by generating point cloud data from stereo camera images. This third dimension allows the system to distinguish between plants that appear to overlap in 2-D images but are actually at different depths, improving row boundary detection accuracy when plants extend over adjacent rows or form gaps within rows.
3Reliability
If 3-D camera systems are used for navigation, then GPS obstruction problems are solved, but the system requires complex processing of image data to maintain accuracy
Solution Approach 1:
The system performs preliminary processing of image data by extracting key features (row line parameters, plant position data) before full navigation computation. The controller pre-processes sequential images to identify and track row structures in advance, creating a simplified representation that reduces the computational complexity of subsequent navigation calculations while maintaining navigation reliability.
Data Source
AI summary
A control system uses visual odometry (VO) data to identify a position of the vehicle while moving along a path next to the row and to detect the vehicle reaching an end of the row. The control system can also use the VO image to turn the vehicle around from a first position at the end of the row to a second position at a start of another row. The control system may detect an end of row based on 3-D image data, VO data, and GNSS data. The control system also may adjust the VO data so the end of row detected from the VO data corresponds with the end of row location identified with the GNSS data.