Object State Identification Using Mixed-Region Recurrent Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting objects in images, particularly those involving machine learning methods like deep neural networks, fail to accurately identify the state of objects represented in images due to issues with time-varying changes in outward appearance, such as mixed states where an object region includes another object.
Innovation Solution
An apparatus and method that utilize a first classifier to detect object regions and determine mixing levels, with characteristics from these regions input into a second classifier with a recursive structure to identify the state of objects, updating internal states based on whether the object region is in a mixed state or not, allowing for accurate identification of time-varying changes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a neural network is used to track objects in time series images, then object tracking capability is improved, but the ability to correctly identify object states with time-varying appearances deteriorates
Solution Approach 1:
The patent divides the object detection task into two separate classifiers: a first classifier for detecting object regions and a second classifier for identifying object states. This segmentation allows each classifier to specialize in its specific function, improving both tracking capability and state identification accuracy independently without interfering with each other.
Solution Approach 2:
The patent introduces an intermediary component that calculates mixing levels to determine whether object regions are in mixed states. This intermediary assessment acts as a mediator between the first classifier's detection and the second classifier's state identification, allowing the system to handle time-varying appearances by recognizing when objects are in mixed states and adjusting identification accordingly.
2Speed
If object regions are detected without considering mixed states, then detection speed is improved, but detection accuracy deteriorates due to inclusion of other objects
Solution Approach 1:
The patent applies partial action by selectively processing only those object regions that are determined to be in non-mixed states for state identification. This partial processing approach maintains detection speed by avoiding unnecessary processing of mixed-state regions while ensuring accuracy by focusing computational resources on clear, unambiguous object regions.
Solution Approach 2:
The mixing level calculation serves as an intermediary assessment that bridges rapid detection and accurate state identification. By introducing this intermediate step, the system can quickly filter out mixed-state regions and proceed with accurate state identification only for clean object regions, thus maintaining both speed and accuracy.
3Measurement precision
If a recursive classifier is used to identify object states, then state identification accuracy for time-varying appearances is improved, but computational complexity increases
Solution Approach 1:
The patent segments the computational task into two distinct classifiers with different complexities. The first classifier handles object region detection with relatively simple operations, while the second recursive classifier handles state identification only for non-mixed-state regions. This segmentation reduces overall computational complexity compared to using a single complex recursive classifier for all regions.
Solution Approach 2:
The system applies partial action by using the computationally intensive recursive classifier only for object regions determined to be in non-mixed states, rather than applying it to all detected regions. This selective application significantly reduces computational complexity while maintaining high state identification accuracy for the regions that require it.
Data Source
AI summary
An apparatus for identifying the state of an object inputs time series images into a first classifier to detect an object region including a predetermined object from each image, determines whether the region of each image is in a mixed state in which the region includes another object other than the object, chronologically inputs characteristics obtained from pixel values of the region of each image into a second classifier having a recursive structure, and applies a recursively used internal state of the second classifier stored in a memory to the second classifier, identifying the state of the object involving time-varying changes in outward appearance. The apparatus rejects the latest internal state when the region of each image is in the mixed state. The apparatus updates the internal state stored in the memory with this latest internal state when the region is not in the mixed state.