Design of an Image Analysis Application for Remote Monitoring of Oil Storage Tanks

Customer

The Customer is an international company operating in the oil and gas industry.

Challenge

The Customer needed to ensure remote liquid level monitoring in oil tanks to optimize oil inventory management and detect oil leak.

Solution

After analyzing the Customer’s business needs, ScienceSoft designed the desktop application that worked with images delivered via pre-installed onsite cameras and drones. The application would detect oil tanks, identify their number and determine the liquid level in each tank remotely with the help of machine vision in real life. The image recognition process involved the following stages:

Stage 1 – Image preprocessing.

Stage 2 – Oil tank detection.

Stage 3 – Oil tank number detection.

Stage 4 – Oil tank number recognition.

Stage 5 – Area with liquid detection.

Stage 6 – Area with liquid size detection.

Stage 7 – Liquid level detection.

How the application works

First, the application employs artificial neural networks trained on a sufficient amount of data to distinguish oil tanks from the background on the delivered images. For further analysis, the application uses neural networks reinforced with a filtering algorithm to scan the numbers of the detected oil tanks and recognize the number depicted on them.

To detect the oil level in the tanks, the application utilizes computer vision algorithms to capture and process the images of the internal visible part of each tank, its external visible part and the roof. Then, the segmentation and classification of the oil tanks according to the liquid level is conducted.

Results

With the application that facilitated real-time remote detection of the liquid level in oil tanks, the Customer could optimize the oil inventory management and timely detect oil leaks.

Technologies and Tools

Linux, Python, PyQt, OpenCV, Keras, TensorFlow, NumPy