Identification of defect types in liquid pipelines for classification and computing severity thereof

Patent metadata

Field	Value
Publication number	US-11790518-B2
Application number	US-202117357210-A
Country	US
Kind code	B2
Filing date	Jun 24, 2021
Priority date	Jul 29, 2020
Publication date	Oct 17, 2023
Grant date	Oct 17, 2023

Abstract

Current inspection processes employed for pipeline networks data acquisition aided with manually locating and recording defects/observations, thus leading labor intensive, prone to error and a time-consuming task thereby resulting in process inefficiencies. Embodiments of the present disclosure provide systems and methods for that leverage artificial intelligence/machine learning models and image processing techniques to automate log and data processing, reports and insights generation thereby reduce dependency on manual analysis, improve annual productivity of survey meterage and bring in process and cost efficiencies into overall asset health management for utilities, thereby enhancing accuracy in defect identification, analysis, classification thereof.

First claim

What is claimed is:

1. A processor implemented method, comprising:
receiving, via one or more hardware processors, an input data comprising at least one of a video data, and one or more images from an image capture device, wherein the input data is specific to a liquid pipeline of a transmission pipeline type, a distribution pipeline type;
extracting, via the one or more hardware processors, one or more optimum frames from the input data specific to the liquid pipeline upon estimating parameters including average luminescence level, a contrast level, and a blur index;
dehazing the one or more extracted optimum frames to obtain one or more dehazed images by extracting one or more local properties of opacity and air-light for each pixel in the extracted optimum frames, wherein the extraction of the local properties results in obtaining an opacity map that is smoothened to extract haze free images;
identifying, from the one or more dehazed images, one or more identified liquid banks and generating one or more contours based on the one or more identified liquid banks;
detecting a change in a liquid level from the one or more generated contours;
estimating a pose of the image capturing device based on (i) an angle of intersection, (ii) a segment intersection, and (iii) a generated circle obtained from the one or more generated contours using a visual sensor in confined noisy environment, and the segment intersection is calculated using angle of intersection and the generated circle obtained from one or more generated contours;
identifying a first set of objects in the liquid pipeline using the estimated pose;
identifying one or more defects in the liquid pipeline based on the first set of identified objects; and
classifying the one or more defects into one or more categories, wherein the step of classifying including identifying region of interests (ROIs) and correcting image pose considering rotation of the image capturing device, wherein once the pose of the image capturing device is corrected, (i) diameter of detected or identified ROIs and (ii) projection angle of junction(s) or connection (s) are calculated followed by classifying the ROI as either a junction or a connection using heuristics based on material, size, and orientation of the first set of objects in frame(s).

2. The processor implemented method of claim 1 , wherein the step of detecting a change in a liquid level from the one or more generated contours comprises:
fitting two or more lines in each of the one or more generated contours;
determining one or more intersection points of the two or more lines;
calculating an angle of intersection and generating a circle with radius R using the one or more intersection points of the two or more lines, wherein the circle is a heuristic for diameter of the liquid pipeline and the angle of intersection approximates a segment covered by liquid;
calculating a segment intersection using the angle of intersection and the generated circle, wherein an area of segment of intersection point with the generated circle is calculated and liquid level in percentage is derived; and
detecting the change in the liquid level based on the segment intersection k tracking a percentage of liquid level across multiple frames and if exists a predefined percentage change in the liquid level in the frame with respect to a base frame, then that specific frame is identified and classified as liquid level change.

3. The processor implemented method of claim 1 , wherein the one or more defects comprise at least one of (a) one or more fractures, (b) one or more breaks, and (c) one or more cracks, and wherein the one or more defects are identified as at least one of (a) the one or more fractures, (b) the one or more breaks, and (c) the one or more cracks based on a defect ratio area.

4. The processor implemented method of claim 1 , wherein the one or more defects are identified using a semantic segmentation technique.

5. The processor implemented method of claim 1 , further comprising computing a severity of the one or more identified defects based on a location associated therewith.

6. The processor implemented method of claim 1 , wherein the first set of objects are at least one of a structural object type including junction and connection, and cracks and fractures and a functional object type.

7. The processor implemented method of claim 1 , wherein the one or more defects identified around a wall of the liquid pipeline are classified as one of (i) a longitudinal defect type or (ii) a circumferential defect type.

8. The processor implemented method of claim 1 , wherein the one or more defects identified around one or more junctions of the liquid pipeline are classified as one or more junction types, and wherein the one or more defects identified around the one or more junctions of the liquid pipeline are based on a change in structural position of the one or more junctions, wherein the one or more junction types includes incorrect position, intruding, damaged and blocked.

9. The processor implemented method of claim 1 , further comprising:
detecting, using a classification model, a second set of objects that are different from the first set of objects;
pre-processing the second set of objects to obtain a set of pre-processed objects;
applying a set of domain-based rules on the pre-processed objects to obtain at least a subset of the set of pre-processed objects; and
classifying the subset of the set of pre-processed objects.

10. The processor implemented method of claim 1 , wherein one or more joints are identified in the liquid pipeline based on the extracted one or more optimal frames using a set of filters by:
segmenting wall of the liquid pipeline into one or more regions using a semantic segmentation technique to obtain a plurality of regions of interest;
applying a first pre-processing filter on the plurality of regions of interest to obtain a plurality of pre-processed regions of interest;
applying, (i) a second pre-processing filter and (ii) a thresholding technique on the plurality of pre-processed regions of interest to obtain at least a subset of regions of interests; and
applying a transformation technique on the subset of regions of interests to identify the one or more joints.

11. A system, comprising:
a memory storing instructions;
one or more communication interfaces; and
one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
receive an input data comprising at least one of a video data, and one or more images from an image capture device, wherein the input data is specific to a liquid pipeline of a transmission pipeline type, a distribution pipeline type;
extract one or more optimum frames from the input data specific to the liquid pipeline upon estimating parameters including average luminescence level, a contrast level, and a blur index;
dehaze the one or more extracted optimum frames to obtain one or more dehazed images by extracting one or more local properties of opacity and air-light for each pixel in the extracted optimum frames, wherein the extraction of the local properties results in obtaining an opacity map that is smoothened to extract haze free images;
identify, from the one or more dehazed images, one or more identified liquid banks and generating one or more contours based on the one or more identified liquid banks;
detect a change in a liquid level from the one or more generated contours;
estimate a pose of the image capturing device based on (i) an angle of intersection, (ii) a segment intersection and (iii) a generated circle obtai

Assignees

• Tata Consultancy Services Ltd

Inventors

• Gubbi Lakshminarasimha Jayavardhana Rama
• Rangarajan Mahesh
• Raj Rishin
• Hariharan Anand Vishnu
• Bajpai Vishal
• Dandenahalli Venkatappa Vishwa Chethan
• Mishra Pradeep Kumar
• Jat Gourav Singh
• Mani Meghala
• Shankarappa Gangadhar
• Nair Dinesh Sasidharan
• Lipate Shashank
• Lall Vineet
• Mathew Kavita Sara
• Seemakurthy Karthik
• Purushothaman Balamuralidhar

Classifications

• G06T7/0006 Primary

  using a design-rule based approach · CPC title

• G01N21/8851

  Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges (G01N21/8806 and G01N21/93 - G01N21/95692 take precedence; optical measurement of dimensions G01B11/00; optical scanning G02B26/10; image transformation G06T3/00; computerised image enhancement G06T5/00; image processing per se for flaw detection G06T7/0002) · CPC title

• G06T5/007

  Physics · mapped topic

• G06T5/20

  using local operators · CPC title

• G06T7/11

  Region-based segmentation · CPC title