Image Processing Apparatus Motion Weight Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection methods based solely on motion information struggle with accurate detection when objects are stationary or when background changes occur, such as due to shadows, lighting fluctuations, or camera movement, leading to misrecognition and reduced detection accuracy.
Innovation Solution
An image processing apparatus that inputs time-sequential still images, sets candidate regions for object detection, acquires motion information, calculates a weight for the appropriateness of the motion, and corrects the detection likelihood using this information to improve object detection accuracy for both moving and stationary objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion information is used for object detection, then detection accuracy for moving objects is improved, but detection accuracy for stationary objects deteriorates
Solution Approach 1:
The patent changes the parameter of detection criteria by introducing a threshold mechanism that adapts to motion information. When motion information exceeds a threshold, the system uses motion-based detection; when motion information is below the threshold, the system falls back to shape-based detection. This parameter change allows the system to maintain high detection accuracy for both moving and stationary objects without relying solely on motion information.
2Productivity
If background difference method is used to extract moving regions, then detection speed is improved, but detection accuracy deteriorates when background changes occur
Solution Approach 1:
The patent introduces motion information as an intermediary element that mediates between background difference extraction and final object detection. The motion information acts as a filter that validates whether extracted regions are true objects or false positives caused by background changes. This intermediary approach maintains the speed advantage of background difference method while correcting its accuracy problems through motion-based verification.
3Device complexity
If only shape information from still images is used for detection, then detection simplicity is maintained, but misrecognition occurs due to noise and incidental textures
Solution Approach 1:
The patent merges shape information from still images with motion information from video sequences to create a hybrid detection system. The shape information provides the basic detection framework and simplicity, while the motion information adds the capability to distinguish true objects from false positives. This combination resolves the contradiction by maintaining relative simplicity while significantly improving detection accuracy through the synergistic integration of multiple information sources.
Data Source
AI summary
An image processing apparatus includes an input unit configured to input a plurality of time-sequential still images, a setting unit configured to set, in a still image among the plurality of still images, a candidate region that is a candidate of a region in which an object exists, and to acquire a likelihood of the candidate region, a motion acquisition unit configured to acquire motion information indicating a motion of the object based on the still image and another still image that is time-sequential to the still image, a calculation unit configured to calculate a weight corresponding to an appropriateness of the motion indicated by the motion information as a motion of the object, a correction unit configured to correct the likelihood based on the weight, and a detection unit configured to detect the object from the still image based on the corrected likelihood.