Development of termite detection system based on acoustic and temperature signals Muhammad Achirul Nanda a , Kudang Boro Seminar a,⇑ , Dodi Nandika b , Akhiruddin Maddu c a Department of Agriculture and Biosystem Engineering, Faculty of Agricultural Engineering and Technology, Bogor Agricultural University, Bogor 16680, Indonesia b Department of Forest Products, Faculty of Forestry, Bogor Agricultural University, Bogor 16680, Indonesia c Department of Physics, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University, Bogor 16680, Indonesia article info Article history: Received 7 May 2019 Received in revised form 12 July 2019 Accepted 1 August 2019 Available online 5 August 2019 Keywords: Acoustic Temperature Termite detection system Termite signals abstract The subterranean termite, belonging to the genus Coptotermes, is the pest that causes maximum destruc- tion in wooden structures, forest trees and crops. Owing to its aggressiveness and hidden existence, it is difficult to visually determine whether a termite infestation is active and damage is occurring. Consequently, the development of a termite detection system as a critical component of termite pest con- trol systems is necessary. A novel termite detection system was developed based on acoustic and temper- ature signals. The system has two capabilities: it can detect the presence of termites and estimate the population size. In this study, a support vector classification (SVc) and artificial neural network (ANN) algorithm were applied to recognize the termites’ signals. After optimizing various input types, kernel functions and model parameters, a robust model was successfully constructed with a specific capability, i.e., a SVc model for detecting termites and ANN model for estimating the termite population. Based on the performance assessment, the proposed termite detection system can detect the presence of termites with an accuracy of 93.83% and estimate their population with a root mean square error (RMSE) of 123.828. The results of our study indicated that the embedded models in the proposed termite detection system successfully proved the feasibility of detecting the presence of termites and simultaneously esti- mating the size of their population. Ó 2019 Published by Elsevier Ltd. 1. Introduction Termites play a vital role in the natural world. As ecosystem engineers, they help break down and decompose dying plants into food for other animals [1,2]. However, when termites start to con- sume the wood in buildings, residential homes, and other commer- cial products and structures, they are considered to be harmful pests. It is widely reported that termites are the most problematic pest in cultivated crops and trees in forests [3,4]. They have man- aged to attack on Acacia crassicarpa plantation, Riau Province, Indonesia [4]. About 10% of the estimated 4000 termite species (about 2600 taxonomically known) are economically significant as a pest that can cause serious damage to wooden structures [5]. The main actor of this damage (90%) is caused by subterranean termites. In Indonesia, the subterranean termite Coptotermes curvi- gnathus causes extensive damage in some areas and becomes the most dangerous one [6,7]. The control and repair costs incurred as a result of termite infestations in Indonesia have been estimated to be approximately IDR 8.7 trillion annually [6]. As a solution, these costs could be significantly reduced through the develop- ment of termite detection system. Detecting the presence of termites is difficult. All termites spe- cies, including drywood and subterranean termites, lead a cryptic or hidden lifestyle, which prevents them from being discovered [1]. However, when termite infestation is active and building dam- age is occurring, termites produce a variety of distinctive signa- tures, including acoustic signals [5,8–10], temperature [11,12], gases (CO 2 ,H 2 , chloroform and methane) [6,12,13] and moisture content [14]. Fundamentally, existing termite detection systems are currently designed based on those signals. So far, acoustic signal have dominated various sophisticated devices and have been mar- keted such as Termatrac TM and Termatrac T3i (Termatrac TM , Aus- tralia) [15,16], AED-2010L and AED-2000 (Acoustic emission consulting, Fair Oaks, CA) [17], Termite tracker (Dunegan engineer- ing, Midland, TX) [17], Insecto-scope (Sound technologies, Kilgore, TX) [17] and Insect detector (DowAgrosciences) [17,18]. In addition, the temperature-based termite detection devices (Flir thermal imaging, USA) and gas detection devices (CO 2 termite detection, Oak Island) have also been successfully marketed [19,20]. However, all of the sophisticated devices that currently exist have a single focus, i.e., detecting termites. In this study, we propose a nonde- https://doi.org/10.1016/j.measurement.2019.106902 0263-2241/Ó 2019 Published by Elsevier Ltd. ⇑ Corresponding author. E-mail address: seminarkudangboro@gmail.com (K. Boro Seminar). Measurement 147 (2019) 106902 Contents lists available at ScienceDirect Measurement journal homepage: www.elsevier.com/locate/measurement