Abstract
Background: The problems to leaf in plants are very severe and they usually shorten the lifespan of plants. Leaf diseases are mainly caused due to three types of attacks including viral, bacterial or fungal. Diseased leaves reduce the crop production and affect the agricultural economy. Since agriculture plays a vital role in the economy, thus effective mechanism is required to detect the problem in early stages.
Methods: Traditional approaches used for the identification of diseased plants are based on field visits which is time consuming and tedious. In this paper a comparative analysis of machine learning approaches has been presented for the identification of healthy and non-healthy plant leaves. For experimental purpose three different types of plant leaves have been selected namely, cabbage, citrus and sorghum. In order to classify healthy and non-healthy plant leaves color based features such as pixels, statistical features such as mean, standard deviation, min, max and descriptors such as Histogram of Oriented Gradients (HOG) have been used.
Results: 382 images of cabbage, 539 images of citrus and 262 images of sorghum were used as the primary dataset. The 40% data was utilized for testing and 60% were used for training which consisted of both healthy and damaged leaves. The results showed that random forest classifier is the best machine method for classification of healthy and diseased plant leaves.
Conclusion: From the extensive experimentation it is concluded that features such as color information, statistical distribution and histogram of gradients provides sufficient clue for the classification of healthy and non-healthy plants.
Hidayat ur Rahman, Nadeem Jabbar Ch, SanaUllah Manzoor, Fahad Najeeb, Muhammad Yasir Siddique, Rafaqat Alam Khan. (2016) A comparative analysis of machine learning approaches for plant disease identification, Advancements in Life Sciences, Volume 4, Issue 4.
-
Views
1185 -
Downloads
4
