8000 IEEE SENSORS JOURNAL, VOL. 17, NO. 23, DECEMBER 1, 2017 Deployment of a Set of Camera Trap Networks for Wildlife Inventory in Western Amazon Rainforest Luis Camacho , Reynaldo Baquerizo, Joel Palomino, and Michel Zarzosa Abstract— The use of multimedia wireless sensor net- works (MWSN) has spread for many sensing applications. An MWSN was deployed for animal monitoring in the Amazon rainforest of Peru. Tapirduino, the developed trap camera sensor node, is composed by an Arduino-like PCB, a CMOS camera, an IR flash, a PIR sensor, a SD card, and 900-MHz radio. The challenges tackled in this project were to overcome the high attenuation of radio frequency in the middle of the jungle and to reduce energy consumption until achieving a very low level. Index Terms— Arduino, CMOS camera, natural capital, MWSN, western Amazon, wildlife monitoring. I. I NTRODUCTION T HE western Amazon is the biologically richest part of the Amazon basin, unlike the eastern (Brazilian) Amazon, it is still a largely intact ecosystem that faces several threatens for wilderness conservation. Putting in value the rainforest is the major strategy to fight against deforestation, land degrada- tion, unsustainable logging and poaching [1]. Wildlife inventories help to increase the value of the forest by showing the rainforest biodiversity and warning about its preservation especially to decision-makers. Despite this is an outstanding topic, very few wildlife inventories have been made in Loreto, the Amazonian region of Peru, most of them covering small territory. Currently, multimedia wireless sensor networks (MWSN) are used in different scenarios [2]–[4] and due to its char- acteristics have become an excellent tool to study wildlife. In this context, a MWSN called TAPIRnet was developed. The TAPIRnet project has been divided into three parts: 1) The sensor nodes or cameras trap. 2) The access networks that moves the data from the sensor nodes to one or more hub nodes. 3) The backbone or transport network that exports the data from the gateways to the Internet under a store and forward scheme. In this paper, the design and development of the two first goals of TAPIRnet project are described. The trap cam- eras are photographic cameras triggered by motion sensor. Manuscript received March 18, 2017; revised June 4, 2017; accepted September 21, 2017. Date of publication October 6, 2017; date of current version November 10, 2017. This work was supported by the INNOVATE Program of the Peruvian Government under Grant 152-FINCyT-IA-2013 and Grant 123-PNICP-PIAP-2015. The associate editor coordinating the review of this paper and approving it for publication was Prof. Nitaigour P. Mahalik. (Corresponding author: Luis Camacho.) The authors are with the Pontifical Catholic University of Peru, Lima 15088, Peru (e-mail: camacho.l@pucp.pe; reynaldo.baquerizo@pucp.pe; j.palomino@pucp.pe). Digital Object Identifier 10.1109/JSEN.2017.2760254 Several methods are used to monitor fauna, according to the taxonomic group. For example for mammals [5] the combination of mul- tiple techniques, such as registration of footprints or feces and capturing photos by cameras trap are recommended. Compared to other sampling methods, such as direct observa- tion, capture or finding signs; cameras trap are more effective for studying nocturnal, elusive or lonely terrestrial mammal species (those weighing more than 1 kg). The trap cameras cause minimal disturbances in the studied species, are easy to deploy and provide additional information on the distribution and habitat use, population structure and behavior of species. Additionally, photographic records can be reviewed by other researchers. Zoologists have already used commercial cameras trap for inventories [6], [7]. These experiences only partially met the requirements of relevance and sustainability, because these cameras presented some disadvantages: 1) high energy con- sumption, commercial cameras trap can remain turned on con- tinuously just for few hours, 2) cost of each commercial trap camera is not negligible and, in order to accumulate sufficient monitoring data within a reasonable period of time, a large number of cameras trap are required and 3) the lack of appro- priate communication systems among the cameras trap, forcing biologists to walk several kilometers to gather information. Most of commercial cameras trap have not been designed for wildlife inventory but for sport hunting in northern hemi- sphere forestlands and African savannah. The use of more appropriate non-commercial cameras trap have been reported previously in India to monitor tigers’ activity [8], but authors finished their work without testing their design in field. Since commercial cameras have critical disadvantages to be used within the Amazon and no other options are available, it has been necessary to develop an appropriate new camera trap with the following characteristics: fast response time to take pictures, size and cost reduction of each unit, use of IR flash that doesn’t disturb animals, automated operation, self- programming, status information and low energy consumption. Open software and hardware enable local development of more appropriate technology for wildlife monitoring in the Amazon rainforest. Fig. 1 shows how ICT local development could cause positive impact on biodiversity preservation, doing the task much more sustainable. This paper has next sections: II) network design, III) hardware design, IV) laboratory results, V) deployment into rainforest, VI) experimental results and VII) conclusions and future work. 1558-1748 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.