1：Experimental Design and Method Selection:

The study uses a deep convolution neural network (CNN) model with an adaptive incomplete iris preprocessing mechanism. The preprocessing involves normalizing iris images and using inner or outer circle methods based on the level of occlusion. Iris region segmentation is done via line fitting and circle fitting to extract features. The CNN model employs pixel coding for irregular iris regions and uses local receptive fields, weight sharing, and sub-sampling for feature extraction and classification.

2：Sample Selection and Data Sources:

Test samples are selected from the Chinese Academy of Sciences iris library (CASIA), with 50 persons and 40 images per person, totaling 2000 samples. Samples are randomly divided into training and test sets using the sklearn library in Python, with the test set accounting for 10%.

3：0%. List of Experimental Equipment and Materials:

3. List of Experimental Equipment and Materials: The experimental setup includes a computer running Windows 10, Python 3.5, an Intel i5 CPU, and software tools such as TensorFlow for deep learning framework and OpenCV for computer vision. No specific hardware for iris acquisition is detailed.

4：5, an Intel i5 CPU, and software tools such as TensorFlow for deep learning framework and OpenCV for computer vision. No specific hardware for iris acquisition is detailed. Experimental Procedures and Operational Workflow:

4. Experimental Procedures and Operational Workflow: Iris images are preprocessed using the adaptive mechanism (inner or outer circle filling based on occlusion). Preprocessed images are normalized to 128x128 pixels. The CNN model is constructed with three hidden layers, convolution kernels of size 3x3, ReLU activation function, and softmax output for classification. Training uses the Adam optimization algorithm with cross-entropy loss function. Dropout is applied to prevent overfitting. Training involves multiple rounds (e.g., 1000 to 2700 rounds) with monitoring via TensorBoard.

5：Data Analysis Methods:

Accuracy and cross-entropy loss are measured during training and testing. Results are compared with classical algorithms. Statistical analysis includes convergence rounds and training time.