Algorithm Chaining for Object Detection Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current object detection technologies face challenges in efficiently and accurately locating objects in vast geographic areas, such as during the Malaysian Boeing 777-200ER airliner disappearance, due to high costs, limited search time, and inefficient processing of large data sets, with traditional ISR techniques and crowd sourcing methods having drawbacks like accessibility issues and potential cyberattacks.
Innovation Solution
A computer-implemented method and system that selects and orders a subset of algorithms based on cumulative trained probability of correctness and conditional probability to reduce processing load, using a chain of algorithms to selectively cull source data and improve object detection efficiency, with stochastic math models and discrete event simulators to compute confidence levels and optimize algorithm chains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If traditional ISR techniques and crowd sourcing methods are used to locate objects in vast geographic areas, then coverage area is improved, but processing time and resource consumption increase significantly
Solution Approach 1:
The patent segments the vast geographic search area into multiple zones and processes images in parallel across distributed computing resources. The system divides the 24,000 sq. km search area into manageable segments that can be processed simultaneously, reducing overall processing time while maintaining comprehensive coverage.
Solution Approach 2:
The patent implements preliminary filtering and preprocessing of images before full analysis. Automated algorithms perform initial object detection and filtering on incoming images, pre-processing data to identify potential targets of interest before human analysts review them, significantly reducing the time required to process vast numbers of images.
2Measurement precision
If more manpower and technology are applied to identify objects, then detection capability is improved, but processing load and resource consumption increase
Solution Approach 1:
The patent introduces automated algorithms and machine learning models as intermediaries between raw image data and human analyst review. These intermediary systems perform initial detection, filtering, and prioritization of images, reducing the processing load on human analysts while maintaining or improving detection capability through automated pattern recognition.
Solution Approach 2:
The patent replaces manual image review processes with automated computer-based algorithms for initial detection and filtering. This substitution of mechanical human labor with automated systems reduces processing load on human resources while enhancing detection capability through consistent, scalable automated analysis of large image volumes.
3Loss of information
If conventional data processing methods are used for Big Data, then data analysis capability is improved, but processing time and computational resources increase
Solution Approach 1:
The patent applies partial processing by focusing computational resources on analyzing only those images and data elements that contain potential targets of interest, rather than processing all available data uniformly. The system performs selective analysis on subsets of data identified as relevant, reducing overall computational resource consumption while maintaining comprehensive data analysis capability for critical elements.
Data Source
AI summary
A system and method improves the probability of correctly detecting an object from a collection of source data and reduces the processing load. A plurality of algorithms for a given data type are selected and ordered based on a cumulative trained probability of correctness (Pc) that each of the algorithms, which are processed in a chain and conditioned upon the result of the preceding algorithms, produce a correct result and a processing. The algorithms cull the source data to pass forward a reduced subset of source data in which the conditional probability of detecting the object is higher than the a priori probability of the algorithm detecting that same object. The Pc and its confidence interval is suitably computed and displayed for each algorithm and the chain and the final object detection.