Advances in automatic identification of flying insects using optical sensors and machine learning

Abstract

Worldwide, farmers use insecticides to prevent crop damage caused by insect pests, while they also rely on insect pollinators to enhance crop yield and other insect as natural enemies of pests. In order to target pesticides to pests only, farmers must know exactly where and when pests and beneficial insects are present in the field. A promising solution to this problem could be optical sensors combined with machine learning. We obtained around 10,000 records of flying insects found in oilseed rape (Brassica napus) crops, using an optical remote sensor and evaluated three different classification methods for the obtained signals, reaching over 80% accuracy. We demonstrate that it is possible to classify insects in flight, making it possible to optimize the application of insecticides in space and time. This will enable a technological leap in precision agriculture, where focus on prudent and environmentally-sensitive use of pesticides is a top priority.

Detection of insect health with deep learning on near-infrared sensor data

“…Automated insect monitoring sensors are currently used to remotely count insect flights (11), record wing beats (12), and classify insects to species (13,14). The incorporation of optical signals with machine learning algorithms enables increased accuracy in remote insect species classification (13). Here we used autonomous near-infrared sensors to study if backscattered light and wingbeat frequencies can differentiate between infected and healthy insects.…”

“…Automated insect monitoring sensors are currently used to remotely count insect flights (11), record wing beats (12), and classify insects to species (13,14). The incorporation of optical signals with machine learning algorithms enables increased accuracy in remote insect species classification (13).…”

Edge Computing for Vision-Based, Urban-Insects Traps in the Context of Smart Cities

“…In recent years, we are witnessing an upsurge of interest in technologically advanced devices as applied to automatic insect detection, counting, and identification [1]. There are mainly three major approaches: (a) optical counters attached to the entrance of traps that target specific pests using lures (pheromones in the case of lepidoptera [2] and palm pests, soil arthropods [3,4] or scents and CO 2 in the case of mosquitoes [5]), (b) camera-based traps that take a picture of their internal space [6][7][8][9][10][11][12][13][14], and (c) near infrared sensors [15] and lidars that emit light covering a volume of space of the open field and registering the backscattered wingbeat signal of flying insects [16][17][18]. All approaches have advantages and disadvantages depending on the application scenario.…”

Identification of Flying Insects in the Spatial, Spectral, and Time Domains with Focus on Mosquito Imaging

“…Such systems have been employed in many field experiments on different continents, where monitoring of wing-beat frequencies was also performed (see, e.g., [ 30 , 31 , 32 , 33 , 34 ]. Analysis of light depolarization [ 9 ], insect flight speed [ 33 ], and differential back-scattering using two laser wavelengths (see, e.g., [ 35 , 36 , 37 ]) has also been accomplished. Normally, Scheimpflug lidar systems operate with elastic back-scattering from the targets.…”