Key Object Identification in Moving Image Tracing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing techniques fail to distinguish between the intended key object and other objects in a moving image, leading to incorrect selection of representative images and inclusion of irrelevant intervals in digests.
Innovation Solution
An image processing apparatus that includes an input unit, a detection unit, a tracing unit, and a key object determination unit to detect and trace objects in a moving image, determine motion, and identify key objects based on overlapping intervals and camera operation information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional object detection techniques are used to retrieve objects from moving images, then objects can be detected and representative images can be generated, but the system cannot distinguish between key objects (intended captures) and other objects (accidental inclusions), leading to incorrect representative image selection
Solution Approach 1:
The patent segments the object detection process into multiple independent analysis dimensions: detection unit identifies all objects, tracing unit tracks motion patterns, and key object determination unit analyzes camera operations. This segmentation allows each component to specialize in one aspect, collectively achieving accurate key object identification by combining results from multiple independent analyses.
Solution Approach 2:
The patent introduces a tracing unit as an intermediary component between detection and identification. This intermediary tracks motion patterns and provides additional contextual information about object behavior, enabling the system to distinguish key objects from other objects by analyzing motion characteristics alongside visual data.
2Productivity
If representative images are generated from all detected objects, then comprehensive coverage is achieved, but the digest includes irrelevant intervals containing non-key objects, reducing information quality
Solution Approach 1:
The patent performs preliminary analysis of camera operations and object motion patterns before generating the final digest. By pre-identifying key objects through multi-dimensional analysis and pre-determining their time intervals, the system avoids including non-key objects in the digest, ensuring high accuracy while maintaining efficient processing.
Solution Approach 2:
The patent extracts only the relevant information (key objects and their time intervals) from the complete set of detected objects. By selectively extracting key object data based on camera operation analysis and motion tracking, the system generates a precise digest that excludes irrelevant non-key objects, improving both accuracy and efficiency.
3Adaptability or versatility
If the system analyzes multiple object attributes (size, number, position) to select representative images, then selection criteria are established, but these attributes do not indicate whether the object is the intended key object, leading to incorrect selections
Solution Approach 1:
The patent adds a new dimension of analysis by examining camera operation data alongside object attributes. Instead of relying solely on object characteristics like size and position, the system analyzes the temporal and operational context of camera movements, creating a multi-dimensional identification framework that accurately distinguishes key objects from other objects.
Solution Approach 2:
The patent creates a universal identification framework that combines multiple function analyses: object detection, motion tracking, and camera operation interpretation. This multi-functional approach allows the system to identify key objects through diverse criteria, making the identification process robust and adaptable to various scenarios while maintaining high precision.
Data Source
AI summary
A face image is detected for each frame at a predetermined interval in moving image data, and the face image is traced using a frame in which the face image is detected and frames subsequent to the frame. A face sequence including an interval in which the face can be traced and motion velocity vectors of the face indicating a change in the position of the face image in the interval is generated based on the tracing result. Further, camera operation information about when the moving image data is acquired is generated from the frame image of the moving image data. When there is an overlap in the plurality of intervals in which the face images are traced, the face being tracked by the camera is determined using the face sequence and the camera operation information of each of the plurality of face images. The face determined to be tracked is then determined to be a key object.