Cognition-Inspired Computational Methods for Automated Chikoo Plant Disease Diagnosis: A Review of Image Processing and Deep Learning

Abstract

The problem of precision in early diagnosis of Chikoo (Manilkara achras) plant diseases remains a serious lapse in digital plant pathology, although significant advances in the computation diagnosis of other horticultural and field crops have been made. This review includes the development of automated plant disease detection- classically image processing or machine-learning pipelines up to deep learning structures, transfer learning procedures, as well as emergent cognition-based motivated and explainable systems. The survey is an integration of the evidence of image acquisition studies, feature extraction approaches by human hands, SVM/KNN/RF-based classifiers, CNNs and transformer model, attention, severity-aware networks, and explainable AI, including Grad-CAM, SHAP, and LIME. Whereas similar crops (like tomato, apple, peach, rice, and arecanut) were proven, Chikoo does not have any recorded computational dataset or diagnostic device. Through the awareness of multidisciplinary knowledge, i.e., farmer decision-making behaviour, mobile-based diagnostic services, and geolocation-linked high-throughput, this review presents a Chikoo-friendly research agenda in the future. The paper identifies the necessity of curated Chikoo datasets, transfer-learning baselines, attention-based models, explainability as well as deployable smartphone applications that combine human-like perception and inference. Altogether, this review places cognition-motivated explainable deep learning as a perspective of developing reliable, transparent, and usable by a farmer, Chikoo disease diagnostic frameworks.