Development of automatic pest sampling and detection system for cash crops

• Main Author: Hadi, Mustafa Kareem
• Format: Thesis
• Language: English
• Published: 2019
• Subjects: Agricultural engineering; Pests - Control; Cash crops
• Online Access: http://psasir.upm.edu.my/id/eprint/89902/1/FK%202020%2025%20ir.pdf; http://psasir.upm.edu.my/id/eprint/89902/

Detection and counting insects constitute a significant challenge in the field of agriculture, especially in tropical countries like Malaysia along with some temperate regions. However, among various biotic issues of agricultural production, pest infestation is the major challenge with the warm humid environment surrounding the crops that encourage the existence survival and proliferation of the pests. As a result, agricultural pests are a serious threat to crops and cause substantial decreases in agricultural yield, causing economic losses as well as adversely affecting the economies of several countries, particularly those that are heavily dependent on agriculture. Therefore, the primary objective of this research is the design and development of a prototype for an automatic Pest Sampling and Detection (PSD) system for cash crops (maize, okra, pineapple, and chili). An automatic system was designed as the hardware part for this system to handle the sampling operation. The system consists of an extendable tripod equipped with a vertical arm with a camera attached, rotary sticky box, protection box, and a controller. The process of insect detection and counting is starting with image acquisition, image preprocessing, and morphological operations. Connected components algorithm was implemented for insect detection and counting. This algorithm can be applied by using MATLAB image processing toolbox. Different kernel functions such as disk, diamond, square, and sphere are used as matching functions for insect detection and counting algorithm. The result of testing the hardware system of the automatic system shows its reliability and flexibility to provide accurate movements in two degrees of freedom as well as its dependability and system protection. Besides that, the result of testing the software system with the conducted experiment shows that the highest counting accuracy by the connected component labeling algorithm is 85.2% by using a sphere kernel function. The accuracy of other kernel functions: disk, diamond, and square are 83.8%, 84.4%, and 62.8% respectively. Finally, it can be concluded that the proposed prototype of an automatic pest sampling and detection system can play a significant role in increasing crop productivity and the management of pest insects in agricultural fields.