Autonomous Object Detection for Transparent Surface Occupancy Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional systems using only active sensors fail to reliably detect difficult-to-detect objects such as transparent or reflective surfaces, while passive sensors provide less precise data, leading to uncertainties in occupancy mapping for autonomous mobile devices navigating physical spaces.
Innovation Solution
Combining active and passive sensors, where a neural network processes image data to identify difficult-to-detect objects, and stereovision techniques determine depth data to update the occupancy map, enhancing the precision of object location, shape, and size by integrating probabilities from both sensor types.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only active sensors are used for object detection, then measurement precision is improved, but reliability deteriorates for difficult-to-detect objects such as transparent or reflective surfaces
Solution Approach 1:
The patent combines active sensors (laser range finder, time-of-flight camera) with passive sensors (stereo camera, monocular camera) into an integrated sensor system. The active sensors provide precise depth measurements while passive sensors detect difficult-to-detect objects through image data. The system merges data from both sensor types to achieve reliable and precise object detection, particularly for transparent or reflective surfaces that passive sensors can identify but active sensors may miss.
2Reliability
If passive sensors are used for object detection, then reliability is improved for difficult-to-detect objects, but measurement precision deteriorates
Solution Approach 1:
The system integrates passive sensors (stereo or monocular cameras) with active sensors to compensate for the precision limitations of passive sensors. The passive sensors reliably detect difficult-to-detect objects and provide image data, while the active sensors supply precise depth measurements. The neural network processes image data from passive sensors to identify objects, then the system combines this with active sensor data to achieve both reliability and precision in object detection.
Solution Approach 2:
The neural network acts as an intermediary that processes image data from passive sensors to identify difficult-to-detect objects. It bridges the gap between passive sensor detection capability and active sensor precision by selecting and processing relevant image data, then integrating it with active sensor measurements to produce accurate object detection results.
3Measurement precision
If both active and passive sensors are combined, then reliability and measurement precision are improved, but device complexity increases
Solution Approach 1:
The patent merges multiple sensor types (active and passive) into a unified sensor system that shares common processing infrastructure. The system integrates laser range finders, time-of-flight cameras, stereo cameras, and monocular cameras, processing their data through a unified neural network and occupancy map generation pipeline. This merging approach achieves high reliability and precision while managing complexity through shared processing resources and integrated data flow.
Solution Approach 2:
The sensor system is designed with multi-functionality, where the neural network processes image data from passive sensors and integrates it with active sensor data through a unified occupancy map generation process. The system can detect various object types (including difficult-to-detect transparent or reflective surfaces) and generate comprehensive occupancy maps, achieving multiple detection goals through a single integrated system rather than separate specialized systems.
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 combination allows autonomous mobile devices to safely navigate spaces with improved detection of difficult-to-detect objects, increasing confidence in object presence and reducing the risk of collisions by integrating precise active sensor data with image-based passive sensor data.
Implementation Method 1
Active sensors emit a signal, while passive sensors do not. For most objects within a physical space, active sensors provide precise information about a size, shape, and location of an object.
Implementation Method 2
a passive sensor such as a stereocamera that acquires image data for stereovision does not emit a signal during operation. Instead, stereovision techniques are used to determine information about the size, shape, and location of an object.
Implementation Method 3
A neural network may be trained to identify difficult-to-detect objects based on their appearance in image data.
Data Source
AI summary
An autonomous mobile device (AMD) may move around a physical space while performing tasks. Sensor data is used to determine an occupancy map of the physical space. Some objects within the physical space may be difficult to detect because of characteristics that result in lower confidence in sensor data, such as transparent or reflective objects. To include difficult-to-detect objects in the occupancy map, image data is processed to identify portions of the image that includes features associated with difficult-to-detect objects. Given the portion that possibly includes difficult-to-detect objects, the AMD attempts to determine where in the physical space that portion corresponds to. For example, the AMD may use stereovision to determine the physical area associated with the features depicted in the portion. Objects in that area are included in an occupancy map annotated as objects that should persist unless confirmed to not be within the physical space.