Date Palm Disease Identification Using Features Extraction and Deep Learning Approach

Abstract

The world is advancing in various fields of life, computers are intervening in every task progressively. In this context, biotechnology software applications are on the peak to solve intricate problems in many fields including agriculture. This paper deals with such problems and their time-series severity in the field of date palm diseases. Date fruit, being the richest one in proteins, is also economically important crop for Pakistan. Due to different highly harmful diseases, the crop does not reach the required level of amount and quality as well. Hence, an attempt to identify a common and rather dangerous disease (Sudden Decline Syndrome) is made through image processing techniques. This paper presents a mechanism to identify the mentioned disease at different stages infecting the date palms. The disease is categorically identified so that proper process may also begin to prevent the losses in the yields. The deep learning technique is used to identify the diseases on the basis of texture and color extraction methods. The dataset of 1200 date palm disease images is developed to experiment the proposed mechanism of disease identification and cumulatively achieved the accuracy of 89.4%. Later, statistical analysis is carried out of outcomes to further process through Convolutional Neural Network. This application will benefit at both macro as well as micro level harvesters and stakeholders across the country and globe.