Sensor Network Local Classifiers for Object Tracking

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing surveillance systems face challenges in efficiently tracking moving objects across complex monitoring areas due to the large volume of data required to be transmitted among numerous network nodes, making it difficult to identify and follow objects across multiple sub-areas effectively.

Innovation Solution

The system employs local classifiers trained on relevant objects within each sub-area, transmitting these classifiers instead of object features, allowing for efficient recognition and tracking by limiting the data transmission to only necessary subsets and enabling seamless object identification across adjacent sub-areas.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If object information is transmitted from one network node to another network node for tracking moving objects across sub-areas, then tracking capability is improved, but data transmission complexity and volume increase significantly

Engineering Contradiction:
Improvetracking capabilityVSAvoiddata transmission complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts only the essential classification information needed for tracking from the complete object data set. Instead of transmitting full object information including all features and attributes, the system transmits only the classifier identifiers and essential tracking parameters, significantly reducing data volume while maintaining tracking capability.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent segments the monitoring area into multiple sub-areas, each handled by a local network node with its own trained classifier. This segmentation allows each node to independently process objects in its local area and only exchange necessary information with adjacent nodes, reducing the overall data transmission complexity of the system.

Inventive Principle:
Principle #1Segmentation

2Measurement precision

If all moving object information is transmitted across the network for comprehensive tracking, then tracking accuracy is improved, but transmission time and processing load increase

Engineering Contradiction:
Improvetracking accuracyVSAvoidtransmission time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system extracts and transmits only the minimum necessary information for accurate tracking - specifically classifier identifiers and essential object parameters - rather than transmitting complete object data. This extraction approach maintains tracking accuracy while dramatically reducing transmission time and processing load.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent implements preliminary action by training local classifiers at each network node in advance. These pre-trained classifiers enable rapid local object recognition and classification, so that when objects need to be tracked across node boundaries, only the classification results need to be exchanged, not the raw data, thus reducing transmission time.

Inventive Principle:
Principle #10Preliminary action

3Productivity

If local classifiers are trained for each network node to recognize relevant objects in their sub-area, then recognition efficiency is improved, but system adaptability to new objects decreases

Engineering Contradiction:
Improverecognition efficiencyVSAvoidsystem adaptability
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent implements universality by designing a standardized classifier structure and information exchange protocol that can be applied across all network nodes regardless of their specific sub-area. The classifier framework is universal and can be trained for different object types and environments, allowing the system to maintain high recognition efficiency while adapting to new objects and scenarios through retraining rather than requiring fundamental system changes.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP2189955B1Sensor network system, transfer protocol, method for recognising an object and computer program
Publication Date: 2012.08.29 ROBERT BOSCH GMBH
  • EP2189955B1 patent drawingFigure 1~2
  • EP2189955B1 patent drawingFigure 3

AI summary

The system (1) has a set of network nodes (2) assigned to a partial area of a monitored area, where one of the network nodes has a classification generator. The classification generator is programmably formed to train a local classifier for discrimination of a moving object that is relevant to the partial region. One network node is designed to pass object information e.g. object color, of the moving object to other network nodes whose adjacent partial regions are assigned in the monitored area. The object information is formed as the local classifier. Independent claims are also included for the following: (1) a transmission protocol for transmitting object information from one network node of a sensor network system (2) a method for recognizing an object (3) a computer program with a program code for executing a method for recognizing an object.