A comparative analysis of the hybrid optical neural network-type filters performance within cluttered scenes

We compare the hybrid optical neural network-type of filters' performance within cluttered scenes. We have tested the unconstrained-, constrained-, and modified-hybrid optical neural network filters' tolerance to background clutter in the input scene by the insertion of training images and non-training out-of-class images into different car park scenes. We have plotted the isometrics of the correlation planes for the different conducted tests and we recorded the peak-to-secondary peak ratio values. All the filters have been proven to be able to recognize the true-class objects. However, the comparison for first time of each of the filter's performance in relation with the others demonstrates the benefits of each filter to be employed for recognizing the true-class objects within the cluttered scenes.