An Intelligent Framework for Automated Noise Eradication in the Mango Plant Leaves Images Using Kernel-Based Deep Learning Approach

The world population is growing by 10% annually, which raises the need for agricultural products. However, some of the elements, such as lesser yield, poor quality, and infected crops, greatly annoy farmers, business people, and customers. It has been noted that each year, the aforementioned variables produce a decline in GDP of close to 1%. It has become very important to detect plant diseases as soon as possible because crop disease is one of the most concerning factors for farmers. In order to increase the accuracy of feature extraction and classification, image enhancement is crucial. In the preprocessing, we concentrated on getting rid of the noise that was present in the images of mango plant leaves. Using performance metrics like Mean Square Error, Peak Signal to Noise Ratio, Mean Structural Similarity, Contrast to Noise Ratio, Contrast Improvement index, and normalized Absolute Error, this study compares the limitations of de-noising techniques like spatial and frequency domain filters, fuzzy filters, and de-noising using a convolutional neural network approach. Finding the right combination of noises and filters to eliminate the specific noise required a lot of time-consuming trial and error. We have developed an improved solution to this issue where the Kernel automatically chooses the best and most appropriate combination of filters and noises. Based on the performance, it is found that the Kernel Deep Neural Network technique offers promising outcomes.