Problem Statement
• Traditional paper-based record keeping system: In traditional paper-based record keeping, the files are used for storing data. Sometimes, these files are lost, brunt which led to loss data. So, our project aims to reduce this problem.
• No computerised entry permission: It detects unauthorised entrances.
• Traditional paper-based record keeping system: In traditional paper-based record keeping, the files are used for storing data. Sometimes, these files are lost, brunt which led to loss data. So, our project aims to reduce this problem.
• No computerised entry permission: It detects unauthorised entrances.
Objectives:
The objective of this project is to successfully detect characters of a number plate, segment and recognize them using KNN algorithm.
The objective of this project is to successfully detect characters of a number plate, segment and recognize them using KNN algorithm.
Constraints of the study
• Varying lighting, camera distance and angle, and format of the image possess problem for segmentation and recognition.
• Different fonts, format and low quality makes image segmentation difficult.
• If the image contains too much spoiled license plate or has designs on it, the program can fail to localize the license plate.
• If the license plate happens to be much tilted from horizontal, then the result of segmentation of the license plate is very poor.
• Varying lighting, camera distance and angle, and format of the image possess problem for segmentation and recognition.
• Different fonts, format and low quality makes image segmentation difficult.
• If the image contains too much spoiled license plate or has designs on it, the program can fail to localize the license plate.
• If the license plate happens to be much tilted from horizontal, then the result of segmentation of the license plate is very poor.
Dataset Selection, Preparation and Training
In Supervised Machine Learning Applications, we need a large amount of dataset for the purpose of training. Just as with unsupervised training these training sets specify input signals to the neural network. A labeled data is a data that is similar to the expected input and we know the expected output value for it. In our project, the input is the image of the character and output is the machine-encoded form of the character. In Machine Learning Applications, data set can be obtained by getting samples that best represent the problem domain from the real world or by generating the samples having a close resemblance to the real-world data. Our dataset is of 12x5 pixels.
Techniques/Algorithms:
We use KNN algorithm in this project. And Following are the steps that are conducted during the process:
1. Scanning
2. Pre-processing
3. Segmentation
4. Post-Processing
5. Character Recognition
We use KNN algorithm in this project. And Following are the steps that are conducted during the process:
1. Scanning
2. Pre-processing
3. Segmentation
4. Post-Processing
5. Character Recognition
CONCLUSION AND FUTURE ENHANCEMENTS
Conclusion:
Since, we can’t build such a system which provides 100% accuracy. And On this Number Plate Recognition system we used 16 samples out of which our system could not detect and recognize 4 number plates. And this system detected and recognized 12 number plates successfully. So, our system accuracy of recognition and detection is 75%.
Conclusion:
Since, we can’t build such a system which provides 100% accuracy. And On this Number Plate Recognition system we used 16 samples out of which our system could not detect and recognize 4 number plates. And this system detected and recognized 12 number plates successfully. So, our system accuracy of recognition and detection is 75%.
Limitations:
• It cannot filter out all the noises that exist in the original image.
• Too much blur image can’t be detected.
Future Enhancements:
• Recognize the text of a number plate of different fonts.
• Detection and Recognition with advanced filtering and noise removal and reduction techniques.
• It cannot filter out all the noises that exist in the original image.
• Too much blur image can’t be detected.
Future Enhancements:
• Recognize the text of a number plate of different fonts.
• Detection and Recognition with advanced filtering and noise removal and reduction techniques.
If you want to see full paper then contact me via email or through other social media.