Plants play a vital role in our lives by providing essential food and protecting us from harmful radiation. Among these plants, tomatoes stand out as a widely cultivated and consumable vegetable. With approximately 160 million tons consumed annually worldwide, tomatoes not only contribute to reducing poverty but also have significant nutritional value [^1^][^2^]. Furthermore, tomatoes possess pharmacological properties that protect against various diseases, such as hypertension, hepatitis, and gingival bleeding [^1^].
However, tomato cultivation faces challenges due to diseases and pests, resulting in significant crop losses for small farmers. Approximately 50% of agricultural output is lost due to these factors [^2^]. Consequently, it is crucial to research effective field crop disease diagnosis methods. Manual identification of pests and pathogens is both inefficient and costly. To address this issue, automated AI image-based solutions are needed to provide farmers with reliable and efficient disease detection tools.
The use of images and image processing technology has proven to be a valuable approach for identifying plant diseases. By processing images, image-based computer vision applications enable accurate and cost-effective disease recognition [^3^]. Detecting diseases early and accurately is essential for minimizing ecological damage, ensuring product quality and quantity, and ultimately impacting a country’s economy [^1^].
To meet the increasing global food demand, agricultural production needs to expand by 70% by 2050 [^2^]. However, relying solely on chemicals, such as fungicides and bactericides, to prevent diseases negatively impacts the agricultural ecosystem. Therefore, more efficient disease classification and detection techniques are necessary to maintain a healthy agro-ecosystem. Advanced technologies like image processing and neural networks can enable the early detection of diseases in tomato plants, which can help reduce plant stress and increase production [^1^].
Traditionally, disease identification is based on farmers’ prior experience and visual inspection of plants, which can be unreliable and subjective [^3^]. Mistakenly classifying a disease and providing incorrect treatment can further damage the plant. Moreover, field visits by domain specialists are costly. Hence, there is a critical need for automated disease detection and classification methods based on images, which can effectively replace manual expertise.
Leaf diseases pose a significant challenge in tomato plantations [^4^], and controlling these diseases is a complex and cost-intensive process [^6^][^7^][^8^][^9^]. Diseases like bacteria, late mildew, leaf spot, tomato mosaic, and yellow curved significantly impact plant growth, leading to reduced product quality and quantity [^10^]. Roughly 80-90% of plant diseases appear on leaves, making early detection crucial [^11^][^12^][^13^][^14^].
Technological advancements have brought forth various solutions for leaf disease recognition. In this article, we will explore these solutions and their applications. Specifically, we aim to highlight methods that enhance lesion visibility, address challenges like illumination shifts, poor image contrast, and image size and form variations. Techniques such as image contrast adjustment, grayscale conversion, image resizing, and filtering are employed as preprocessing operations [^15^][^16^][^17^]. Additionally, we delve into object segmentation within images to identify infected regions accurately [^18^].
Classification of the disease is the subsequent step, whereby the sample is assigned to a specific class. This involves surveying one or more input variables to identify the type of disease. Enhancing classification accuracy remains a significant challenge. Finally, datasets dissimilar to the training set are used to validate the classification model.
In the following sections, we delve deeper into the existing literature, describe the materials, methods, and processes involved, analyze and discuss the results, and ultimately conclude our findings.
