Conduit Video Anomaly Detection for Precise Cross-Bore Location
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional conduit inspections for wastewater and sewage systems are labor-intensive and prone to human error, with manual review of video streams being time-consuming and inefficient in detecting anomalies like fissures, cracks, roots, and cross-bores, and lacking specificity in locating these issues within the conduit.
Innovation Solution
An optical imaging device coupled with a machine learning protocol and a controller that automatically detects anomalies by analyzing video streams, applying alerts, and geolocating them using GPS, thereby reducing human intervention and enhancing detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual review of video streams is performed by operators and technicians, then anomaly detection can be conducted, but inspection time and labor costs increase significantly
Solution Approach 1:
The patent replaces the manual mechanical review process with an automated machine learning system that uses computer vision algorithms to detect anomalies in video streams. The machine learning model processes video frames automatically, substituting human operators and technicians with an automated computational system that maintains detection accuracy while dramatically reducing inspection time
Solution Approach 2:
The system enables self-service anomaly detection where the machine learning model independently analyzes video streams without requiring human intervention for each inspection. The automated system performs detection, classification, and location marking of anomalies autonomously, eliminating the need for operators to manually review every video stream while preserving reliable detection capabilities
2Reliability
If manual review of video streams is performed, then anomaly detection can be conducted, but labor costs increase due to vast amounts of time spent by operators
Solution Approach 1:
The patent replaces the manual mechanical review process with an automated machine learning system that uses computer vision algorithms to detect anomalies in video streams. The machine learning model processes video frames automatically, substituting human operators and technicians with an automated computational system that maintains detection accuracy while dramatically reducing inspection time
Solution Approach 2:
The system changes the operational parameters from manual human review to automated machine learning processing. By transforming the inspection process from labor-intensive manual analysis to computational automated analysis, the system maintains anomaly detection capability while significantly improving inspection efficiency and reducing labor costs
3Loss of information
If conventional travel distance measurement is used, then location data can be obtained, but specificity in locating anomalies within the conduit is insufficient
Solution Approach 1:
The patent adds temporal dimension to the location data by recording the specific time when each anomaly is detected during the video stream analysis. This transforms the single-dimensional travel distance measurement into a multi-dimensional location system that includes both spatial position (travel distance) and temporal information (detection time), enabling precise anomaly localization within the conduit
Solution Approach 2:
The system segments the conduit inspection into discrete detectable events by identifying and marking each anomaly's precise location along the video stream. By dividing the continuous video data into individual anomaly instances with specific location markers, the system provides segmented, precise location information for each anomaly rather than a single aggregate travel distance measurement
Data Source
AI summary
A method for automatically detecting an anomaly inside of a conduit in real-time computing. The method includes steps of: moving an optical imaging device of a system inside of the conduit; viewing at least one anomaly inside of the conduit with the optical imaging device; outputting a video stream by the optical imaging device with the at least one anomaly to a user interface of the system; executing an anomaly detection program, by a controller of the system, from a computer readable medium in response to the at least one anomaly being viewed by the optical imaging device from the live video stream. The controller is caused to: automatically detect the at least one anomaly with a machine learning protocol of the anomaly detection program; output the at least one detected anomaly to the control interface; and automatically indicate the at least one detected anomaly on the live video stream.