"卷积核大小 3*3，卷积核移动步长 1，卷积核个数 16，激活函数 ReLU，使用 batch_normal 和 weight_decay"的代码

以下是使用TensorFlow实现上述卷积神经网络的代码：

import tensorflow as tf

# 输入数据的维度
input_shape = (None, 28, 28, 1)

# 定义输入数据的占位符
inputs = tf.placeholder(tf.float32, shape=input_shape)

# 定义卷积核的权重和偏置
conv_weights = tf.get_variable("conv_weights", shape=[3, 3, 1, 16], initializer=tf.truncated_normal_initializer(stddev=0.1))
conv_bias = tf.get_variable("conv_bias", shape=[16], initializer=tf.zeros_initializer())

# 定义卷积层
conv_output = tf.nn.conv2d(inputs, conv_weights, strides=[1, 1, 1, 1], padding="SAME")
conv_output = tf.nn.bias_add(conv_output, conv_bias)

# 定义 batch normalization
conv_output = tf.layers.batch_normalization(conv_output)

# 定义 ReLU 激活函数
conv_output = tf.nn.relu(conv_output)

# 定义权重衰减项
weight_decay = tf.multiply(tf.nn.l2_loss(conv_weights), 0.0001)

# 定义损失函数
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=logits))
loss = tf.add(loss, weight_decay)

# 定义优化器
optimizer = tf.train.AdamOptimizer(learning_rate=0.001)

# 定义训练操作
train_op = optimizer.minimize(loss)