Integration of Computer Vision and Soil Sensor Analytics for Precision Fertigation in Smart Horticulture

Abstract

Smart agriculture is a viable solution to enhance the farming output, exploitation of resources, and sustainability of the environment via smart monitoring and automated crop control methods. Accurate application of water and nutrients based on the needs of crops is an important aspect of contemporary horticulture because it can effectively be done through precision fertigation. However, traditional methods of fertigation are characterized by poor management of resources, delayed stress identification and poor real-time decision making abilities. This study seeks to overcome these difficulties by presenting a holistic intelligent fertigation system that combines the computer vision and soil sensor analytics-based approaches to smart horticulture solutions. The suggested system employs AI-based picture examination to track the health of harvests, detect the indications of stress in plants, and assess visual development traits, whereas soil sensors collect the assessments of environmental and soil parameters (moisture, temperature, humidity, pH, and nutrient amounts) continuously. An automated crop condition classification and provided precision decision support based on a Convolutional Neural Network (CNN)-based computer vision model were used. The gathered sensor and image data was then handled with machine learning algorithms to maximize irrigation and nutrient management plans. The effectiveness of the proposed framework was evaluated through experimental evaluation based on metrics of computer vision performance, such as accuracy, precision, recall, and F1-score. The findings showed better crop surveillance efficiency, easy to control fertigation, minimized water and fertilizer losses, and increased crop yield. The suggested intelligent system is a step towards sustainable horticulture, as it will allow real-time precision agriculture, resource management and automated fertigation management using data.