Description

Question 3 – Image Processing

Programming Question (6 marks totally)   

Remove all watermark and shadow of following image of a cat.  

1.      Submit your bug-free source code with name as ‘watermark_your first name.py’ or ‘watermark_your first name.ipynb’ to Avenue-to-Learn together with this WORD document. (3 marks) 

2.      Here, a sample image is posted with removed watermark/shadow. Now, please put your own post-processed image in the following space (3 marks):

Question 4 – Artificial Neural Network (ANN)

Programming question for a real-world problem (only 4 marks totally) 

Background:

Suppose we were given the task of design a flame detection algorithm based on ANN for a Toxic Waste Incinerator. The intense heat of the fire is intended to neutralize the toxicity of the waste introduced into the incinerator. Such combustion-based technique is commonly used to neutralize medical waste which may be infected with deadly viruses or bacteria.

System Overview:

The figurer below outlines a typical incineration system. As long as a flame is maintained in the incinerator, it is safe to neutralize the waste. However, if the flame were to be extinguished, it would be unsafe to continue to add waste to the combustion chamber as it would exit the exhaust ‘un-neutralized’. Therefore, the system required flame detection sensors to ensure waste neutralization.

Figure. Toxic waste incinerator system overview

Sensor Selection:

Several different flame-detection technologies exist: optical (light), thermal (high temperature), and electrical conduction (ionized particles in the path of the flame). Each system has its advantages and limitations. Due to the hazardous nature of the waste, it is typical that the flame detection stage be duplicated (multi sensors) to avoid any toxic emission when a failure of a single sensor occurs. Each sensor produces binary signal (0 if no flame, 1 if flame is detected). Following figure shows the locations of the three sensors in the toxic waste incinerator system. 

Figure. Toxic waste incinerator with sensor detection

To maximize safety, we would design an algorithm to open the waste valve if all sensors indicate flame; however, it makes the system very susceptible to sensor failures of the opposite kind. Suppose that one of the three sensors were to fail in such a way that it indicated no flame when there was in fact in the incinerator’s combustion chamber. The single failure would unnecessarily shut off the waste vale, resulting in lost production time and waste fuel.

It would be ideal to have a decision-making system that would allow for this type of failure without shutting down the system and still provide sensor redundancy so as to maintain safety in the event that any single sensor failed. A strategy that would meet both needs would be a ‘two out of three’ approach, whereby the waste value is opened if at least two out three sensors indicate a ‘valid’ flame status. Follow table outlines the relationship of sensor inputs and valve output for designing such a decision-making system based on Artificial Neural Network.

Table. Relationship of sensor inputs and valve output

Task: Submit your bug-free source code with name as ‘ANN_your first name.py’ or ‘ANN_your first name.ipynb’ to Avenue-to-Learn together with this WORD document. (4 marks)