Pipeline defect intelligent detection method based on image processing and deep learning, and application thereof
A deep learning and intelligent detection technology, applied in the field of Internet of Things and artificial intelligence, can solve problems such as difficult to distinguish steel pipe defects, inability to detect corrosive defects, poor detection and training effects, etc.
Patent Information
- Authority / Receiving Office
- CN · China
- Current Assignee / Owner
- Publication Date
- 2021-09-10
- Estimated Expiration
- Not applicable · inactive patent
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The invention relates to the technical fields of Internet of Things and artificial intelligence. Background technique
[0002] Pipelines are commonly used liquid and gas transmission media in the petroleum field. Due to the complexity of the working environment and transmission materials, they are prone to corrosion, blockage and even rupture during use. Therefore, the pipelines are regularly inspected to ensure their service life. And security is a must. The traditional manual detection method for pipeline defects is time-consuming and prone to false detection and missed detection. Therefore, an intelligent defect detection method is a better choice.
[0003] The existing intelligent detection methods for oil pipeline defects involve computer vision technology and embedded hardware technology. In terms of hardware, cameras and hardware platforms that can sense the environment, analyze scenes, and make corresponding responses are required. However, d...
Examples
Embodiment 1
[0112] The defect detection model is obtained by the following process:
[0113] data collection
[0114] In the present invention, the data consists of two parts: self-made data set and NEU public data set. The self-made data set comes from a video data set of an oil mining site, and some of its video images are as follows: Figure 4 shown. Since the collected on-site data set mainly comes from short-term video, the data has certain regularity. In order to avoid this problem, the collected video is extracted by frame to obtain image data.
[0115] Because the invention is used to detect whether there are defects in oil pipelines, in order to achieve better detection results, NEU public data sets are added to supplement data diversity. The original label classification of the public data set is: rolling scale (RS), plaque (Pa), cracking (Cr), pitting surface (PS), inclusions (Is) and scratches (Sc), a total of 6 categories Different types of surface defects.
[0116] data...