The proposed system utilizes image processing algorithms to make it easier for visually impaired people to recognize banknotes. The latest banknotes in Bangladesh have blind embossing or blind dots, which can effectively identify the value of the note by touch. As the embossing disappears on long-used banknotes, detecting the true value of the banknote using image processing algorithms could be considered a challenging task.
19 Table 4.3.1: Results using different methods in Ideal Perspective 20 Table 4.3.2: Results using different methods in Distortion Perspective 20. With these statistics it should be mentioned that the increase in the number of visually impaired employees in the government and private work sectors is highly impressive. But one of the main problems faced by a visually impaired person is identifying paper currency due to the similarity of paper size and texture between different banknotes.
Therefore, we proposed a real-time system that will help them recognize currencies and solve this crisis so that visually impaired people feel confident in financial transactions without depending on others. Camera-based systems can be developed using modern image processing techniques, and every smartphone has camera features, making it accessible to the visually impaired. If such a system can be developed, visually impaired people will be able to trust currency recognition.
It is very impressive to see the increase in the number of visually impaired workers in government and private services.
Rationale of the Study
Research Questions
Which image detection approach will be followed as a system will be implemented in an Android application. How will we achieve higher accuracy as the system will be used in financial relationships.
Expected Outcomes
Report Layout
Background
- Introduction
- Related Works
- Research Summary
- Scope of the Problem
- Challenges
The experimental analysis shows satisfactory results as the system recognizes the paper currency with a high rate and fast recognition speed. The idea of developing this kind of systems of visually impaired persons is missing in the literature. Therefore, the need for a real-time system for visually impaired persons is quite high and in this paper we try to mitigate the gap in the literature.
Since the system needs to recognize paper coins in real time, we need to reduce the time complexity for recognition. As the implementation of CNN directly in the Android application is not flexible enough at the moment, we will have to focus on feature-based detection as well as reducing the time complexity. This study focuses on finding a way to develop a real-time Bangladesh currency detection system for visually impaired people.
Recognize banknotes: They will be able to recognize paper currencies with a convenient tool, as the system will be implemented as an Android application running on lower budget smartphones. Participation in microfinance: Visually impaired people will be able to set up their small microfinance business or participate in some jobs created by our government for disabled people. To develop such a system for the visually impaired, we need to create a reliable system with higher accuracy and ease of use.
Usability of Smartphones: Since people with visual impairment face vision problems, operating a smartphone will be impossible for them. To overcome these difficulties we must create such an approach or use a pre-built system that will help them control a smartphone. Time Complexity: Since the target system is a real-time detection system, we should focus on reducing the detection processing time with the highest priority.
Lower camera resolution: 80% of visually impaired people in our country live in rural areas, as well as their financial conditions are not good enough to buy a smartphone that has a good camera. So the system should be developed in such a way that the camera resolution can be moderate. Android App does not support python programming language directly, so the system must be created using programming language JAVA or Kotlin, where there are not enough resources available in this area.
Research Methodology
- Introduction
- Research Subject and Instrumentation .1 Reference Image Database Setup
- Image Preprocessing
- Image Analysis
- Image Recognition
- Recognition Analysis
- Data Collection Procedure
- Statistical Analysis
- Implementation Requirements
- Introduction
- Experimental Results
- Descriptive Analysis
- Summary
Detected description of input image will be saved and system will start matching in Image Recognition part. After input image analysis, the system will now match the descriptors of training images that we generated from Preprocessing to the descriptor of input image. The description of input image will match the training descriptions one by one and calculate the corresponding results.
Each matching result will be saved for the analysis part, where the system will analyze between the results to return the most accurate matching banknote. First, we calculate the sum of corresponding points of all training images of each category of banknotes. Since the system will measure the maximum matching set, chances of inaccuracy will be dramatically reduced.
The system contains a total of 40 training images and was tested in two different perspective images, Ideal and Distortion. The implementation of the system is very linear since we implemented the recognition system as an Android application. The system will be used by people who may not position the camera correctly over the paper notes; as a result, recognition may fail in some cases.
Since the system has to work in real time, we need to focus on the time required to recognize banknotes. The real-time implementation of the system is demonstrated in [13] while practically using the mobile application. By using the ORB detector, descriptors and brute force hamming matcher, the training time (1.12 s) and average matching time for the ideal case is very minimal as it shows 100% accuracy.
Summary, Conclusion, Recommendation and Implication for Future Research
- Summary of the Study
- Conclusions
- Recommendations
- Implication for Further Study
There are some recommendations that should be followed if this research is to be conducted again in the future. For application development Koltin programming language should be used to get faster processing speed. There are many people in the world who are not normal; they face difficulties in their life.
There are other people who may have problems listening, walking, taking and so on.
APPENDIX A Research Reflection
APPENDIX B Related Issues
34;A Classification Method for the Dirty Factor of Banknotes Based on Neural Network with Sine Basis Functions International Conference on Intelligent Computing. 34;Paper Currency Recognition Using Gaussian Mixture Models Based on Structural Risk Minimization International Conference on Machine.
