Efficient object detection technique using DWT based on deep learning

Abstract

Artificial intelligence has great merit in solving many problems, such as the problem of distinguishing objects, distinguishing faces, and vehicles to determine vehicle plate numbers. With the help of image enhancement through noise reduction and image compression, the deep learning process is improved and better results are achieved. This is done by deriving new discrete wavelets from polynomials, especially Hermite polynomials (HP), to arrive at the discrete Hermite wavelet transform (DHWT) by relying on the mother wavelet and arriving at a new Hermite filter to be analyzed. Input image, denoising and image compression, with the help of the new filter, the most important quality standards were achieved, the most important of which were the amount of error, signal noise, number of pixels, and amount of compression. The new filter was used in the convolution process to build the convolutional neural network to complete the deep learning phase with the new wavelets to achieve improvement so that the new and fast algorithm was built to detect objects where the best results were achieved with 98.44% accuracy in 30 seconds.