Working Machine Shadow Profile Extraction for Obstacle Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The complexity of a working machine's shadow on uneven terrain poses challenges for obstacle detection systems, leading to false recognition and decreased efficiency, as existing methods require extensive data storage and processing to distinguish the machine's shadow from obstacles, especially in off-road environments where shadows dynamically change.
Innovation Solution
A surroundings monitoring system that uses a monocular camera and image processing units to extract characteristic patterns, excluding the shadow profile from the obstacle detection target, allowing for accurate detection of objects within the shadow without relying on pre-stored shadow boundary patterns or color analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a plurality of shadow boundary patterns are prepared in advance and stored in a memory apparatus, then shadow recognition accuracy may be improved, but device complexity and initial cost increase significantly
Solution Approach 1:
The patent extracts characteristic amounts (features) from the working machine image in advance and stores only these extracted features along with their corresponding ground regions, rather than storing complete shadow boundary patterns. This preliminary extraction action reduces the data storage requirement while maintaining recognition accuracy.
Solution Approach 2:
The patent extracts only the essential characteristic amounts (features) from the complex shadow boundary patterns and stores these extracted features. This extraction process removes unnecessary data while preserving the key information needed for accurate shadow recognition, thereby reducing device complexity.
2Measurement precision
If countless shadow boundary patterns are prepared for every profile and road surface shape, then recognition accuracy may be improved, but loss of time and processing complexity increase
Solution Approach 1:
The patent changes the representation parameters from complete shadow boundary patterns to extracted characteristic amounts (features). This parameter transformation reduces the data dimensionality and enables faster processing while maintaining recognition accuracy, thus reducing loss of time.
Solution Approach 2:
Instead of storing and processing numerous complete shadow boundary patterns, the patent creates a simplified copy representation using only the essential characteristic amounts extracted from the images. This copying approach reduces processing complexity and time while preserving the necessary recognition information.
3Area of stationary object
If the camera is installed at a high position with great depression angle to monitor surroundings of the huge working machine, then monitoring coverage is improved, but shadow frequency and detection difficulty increase
Solution Approach 1:
The patent replaces the traditional mechanical approach of avoiding shadows by adjusting camera position with an image processing approach. By extracting characteristic amounts and comparing them with stored features, the system can accurately detect objects even when they appear in shadows, thus substituting mechanical constraints with computational solutions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A surroundings monitoring system for a working machine is provided by which, even if an image obtained by imaging surroundings of the working machine includes an own vehicle shadow of a complicated shape, it can be prevented that the presence of the own vehicle shadow has an influence on detection of an object existing around the working machine. The surroundings monitoring system for a working machine includes: a monocular camera 6 that picks up an image of the surroundings of the working machine; a characteristic pattern extraction unit 170 that extracts characteristic patterns in the image based on a characteristic amount of the image; a shadow profile extraction unit 40 that extracts a profile of a region, which can be regarded as a shadow of the working machine in the image, based on the characteristic amount of the image; and an object detection unit 180 that detects an obstacle existing around the working machine based on the remaining characteristic patterns obtained by excluding a shadow profile characteristic patterns positioned on the profile extracted by the shadow profile extraction unit from the characteristic patterns extracted by the characteristic pattern extraction unit.