Performance Analysis of Fingerprint Minutiae Extraction Using Graph and Dual Networks

 

Implementation plan:
*************************

Proposed Dual CNN for Minutiae Extraction
*************************************************

Step 1: Initially, we collect and load data from FVC2000,FVC2002 and FVC2004 Fingerprint Dataset

Step 2: Then, we preprocess the images using normalization, ridge enhancement, binarization, thinning, ROI quality estimation, and quality index mapping.

Step 3: Next, we extract candidate minutiae points using Fingernet, DeepPrint and Spectral techniques.

Step 4: Next, we develop a Dual CNN model for Minutiae Extraction consisting of two parallel networks: Clean Ridge CNN for high-quality regions and Degraded Ridge CNN for low-quality regions. These networks learn pixel-level features using quality indices, and their outputs are combined through a weighted fusion layer to perform feature-level summation. The fused features are used to accurately mark the minutiae points, and displayed as output.

Step 5: Finally, we plot performance metrics for the following

5.1: Number of epochs vs. Accuracy (%)
5.2: Number of epochs vs. Precision (%)
5.3: Number of epochs vs. Recall (%)
5.4: Number of epochs vs. F1-score(%)
5.5: Number of epochs vs. Localisation Accuracy (%)
5.6: Number of epochs vs. Orientation Error (%)
5.7: Number of epochs vs. Detection Rate with Tolerance (%)
5.8: Number of epochs vs. False Minutiae Density (%)
5.9: Number of epochs vs. Missed Minutiae Density (%)
5.10: Number of epochs vs. Hausdorff Distance (pixels)
5.11: Number of epochs vs. Minutiae Pairing Accuracy (%)
5.12: Number of epochs vs. ROC Curve (%)

 

Existing Graph Based CNN for Minutiae Extraction
*******************************************************

Step 1: Initially, we collect and load data from FVC2000,FVC2002 and FVC2004 Fingerprint Dataset

Step 2: Then, we preprocess the images using normalization and ridge enhancement,

Step 3: Next, we extract candidate minutiae points using Fingernet, DeepPrint and Spectral techniques.

Step 4: Next, we develop a Graph-Based CNN for Minutiae Extraction by constructing a graph where nodes represent ridge endings and bifurcations, and edges represent ridge connections. The graph neural network is then trained to identify and mark true ridges and bifurcation points, and the extracted minutiae are displayed as output.

Step 5: Finally, we plot performance metrics for the following

5.1: Number of epochs vs. Accuracy (%)
5.2: Number of epochs vs. Precision (%)
5.3: Number of epochs vs. Recall (%)
5.4: Number of epochs vs. F1-score(%)
5.5: Number of epochs vs. Localisation Accuracy (%)
5.6: Number of epochs vs. Orientation Error (%)
5.7: Number of epochs vs. Detection Rate with Tolerance (%)
5.8: Number of epochs vs. False Minutiae Density (%)
5.9: Number of epochs vs. Missed Minutiae Density (%)
5.10: Number of epochs vs. Hausdorff Distance (pixels)
5.11: Number of epochs vs. Minutiae Pairing Accuracy (%)
5.12: Number of epochs vs. ROC Curve (%)

 

Software Requirements:
*************************

1) Development Tool: MatlabR2023a or above
2)Operating System: Windows 10 (64-bit) or above

 

Dataset Links:
****************
FVC2000 : https://www.kaggle.com/datasets/peace1019/fingerprint-dataset-for-fvc2000-db4-b
FVC2002 Dataset: https://www.kaggle.com/datasets/nageshsingh/fvc2002-fingerprints
FVC2004 Dataset: http://bias.csr.unibo.it/fvc2004/download.asp

Note:
*******

 

1) If the proposed plan does not fully align with your requirements, please provide all necessary details—including steps, parameters, models, and expected outcomes—in advance. Kindly ensure that any missing configurations or specifications are clearly outlined in the plan before confirming.

2) If there’s no built-in solution for what the project needs, we can always turn to reference models, customize our own, different math models or write the code ourselves to fulfil the process.

3) If the plan satisfies your requirement, Please confirm with us.

4) Project based on Simulation only.

5) If you have any dataset to change,kindly provide us before implementing it.