Filter design for small target detection on infrared imagery using normalized-cross-correlation layer in neural networks

A filter design method for a small target detection on an infrared imagery using a normalized-cross-correlation layer in neural networks, comprising steps of;

• normalizing input patches and filter coefficients of convolutional neural network layers for providing a faster convergence in a limited database,
• defining a forward propagation function of a normalization layer in a convolutional neural network,
• defining a derivative function of the normalization layer for a back propagation, wherein the back propagation is used in a neural network training phase,
• training a created neural network by employing a back propagation method and using datasets, wherein the datasets consist of target and background views, and
• using trained neural networks in a small target detection task, wherein a forward function of the normalized-cross-correlation layer is a convolution of a normalized 2D discrete input signal p and layer coefficients filter f, and wherein the normalized 2D discrete input signal p is calculated by the following formula:
• p
• _
• =
• 1
• n
• -
• 1
• ·
• p
• -
• μ
• p
• σ
• p
• where μp represents a mean pixel value and σp represents a standard deviation of a patch p.