Multi-Viewpoint Surveillance Object Depth and Texture Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current surveillance systems face challenges in accurately determining events in a surveillance area due to difficulties in distinguishing between moving objects with varying depths and textures from 2D images, leading to reduced reliability and performance.
Innovation Solution
An image processing apparatus and method that detects moving objects from multiple viewpoints, determines their depths, combines objects with similar depths, and assesses texture similarity to accurately identify events by setting threshold values for depth and texture differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth determination based on disparity vectors is implemented, then measurement precision of object depth is improved, but device complexity increases
Solution Approach 1:
The system performs preliminary action by detecting moving objects and determining their depths before event detection. The depth determination unit calculates disparity vectors for each detected object to establish depth information in advance, which then serves as a basis for subsequent event detection and object combination operations.
Solution Approach 2:
The system applies segmentation by processing each detected object individually through the depth determination unit. Each object's depth is calculated separately using disparity vectors, and objects are then combined or separated based on depth differences and texture similarity, dividing the complex processing into manageable per-object operations.
2Device complexity
If multiple objects with similar depths are combined into one, then device complexity is reduced, but measurement precision of object identification deteriorates
Solution Approach 1:
The system applies local quality by evaluating both depth similarity and texture similarity for each pair of objects before combining them. Objects are combined only if they satisfy both conditions (depth difference < first threshold AND texture difference < second threshold), ensuring that local texture characteristics are preserved as a quality criterion for combination decisions.
Solution Approach 2:
The system changes parameters by using multiple criteria (depth difference and texture difference) with adjustable threshold values to determine object combination. The first threshold value controls depth-based combination while the second threshold value controls texture-based combination, allowing flexible parameter adjustment to balance complexity reduction with identification precision.
3Measurement precision
If texture information analysis is performed to distinguish objects, then measurement precision of object distinction is improved, but use of energy increases
Solution Approach 1:
The system performs partial action by analyzing texture information only for objects that have similar depths. The object combining unit first filters objects based on depth difference < first threshold, then performs texture analysis only on this subset. This partial application of texture analysis reduces overall energy consumption while maintaining distinction precision where it matters most.
4Reliability
If event detection based on depth conditions is implemented, then reliability of event detection is improved, but device complexity increases
Solution Approach 1:
The system performs preliminary action by determining depth information for all detected objects before event detection. The depth determination unit calculates depths in advance, and the event detecting unit then uses this pre-computed depth information to evaluate depth conditions, avoiding redundant depth calculations during event detection and improving overall system efficiency.
Data Source
AI summary
An image processing apparatus such as a surveillance apparatus and method thereof are provided. The image processing apparatus includes: an object detecting unit which detects a plurality of moving objects from at least one of two or more images obtained by photographing a surveillance area from two or more view points, respectively; a depth determination unit which determines depths of the moving objects based on the two or more images, wherein the depth determination unit determines the moving objects as different objects if the moving objects have different depths.