Agricultural Obstacle Detection Using Sensor-Guided Image Trimming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing obstacle detection systems for agricultural work vehicles require high-performance arithmetic units due to the large image size and extensive arithmetic operations needed for detecting obstacles far from the camera, resulting in high costs.
Innovation Solution
An obstacle detection system that includes an obstacle sensor, an obstacle estimation unit, an image capturing unit, an image preprocessing unit, and an obstacle detection unit. The system uses a trimmed image, reduced in size based on obstacle present region information and shooting-angle-of-view information, to reduce the arithmetic load on the obstacle detection unit, thereby improving cost performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a captured image with large image size is used as input image for obstacle detection, then the obstacle detection accuracy is improved, but the arithmetic operation load increases and high-performance arithmetic units are required
Solution Approach 1:
The captured image is divided into multiple regions based on distance from the camera. The image processing is segmented into different processing levels: full-resolution processing for near regions and downsampled processing for far regions. This segmentation allows accurate detection where needed while reducing arithmetic load in less critical areas.
Solution Approach 2:
Different processing qualities are applied to different regions of the image. Regions closer to the camera (where obstacles appear larger and more critical) receive full-resolution processing, while regions farther away receive downsampled processing. This local quality approach maintains detection accuracy for critical areas while reducing overall computational burden.
2Reliability
If the full captured image is processed for obstacle detection, then all obstacles are detected, but the processing time increases and cost performance deteriorates
Solution Approach 1:
The system performs preliminary downscaling of distant regions of the image before obstacle detection processing. By pre-reducing the resolution of far regions, the subsequent obstacle detection algorithm processes fewer pixels, significantly reducing processing time while maintaining adequate detection capability for distant obstacles.
Solution Approach 2:
The system applies partial processing to the image by using full resolution only for near regions and downsampled resolution for far regions. This partial action approach processes only the necessary amount of data at full resolution, achieving acceptable detection completeness with reduced processing time compared to processing the entire image at full resolution.
Data Source
AI summary
An obstacle detection system for an agricultural work vehicle includes: an obstacle estimation unit that estimates a region in a field based on a detection signal from an obstacle sensor in which region an obstacle is present, and outputs obstacle present region information; an image capturing unit that outputs a captured image of the field; an obstacle detection unit that detects the obstacle from an input image and outputs obstacle detection information; and an image preprocessing unit that generates, as an image to be input to the obstacle detection unit, a trimmed image obtained by trimming the captured image so as to include the region in which the obstacle is present, based on the obstacle present region information and shooting-angle-of-view information regarding the image capturing unit.