Insects monitoring using a triggered camera device for existing traps

Recent studies have highlighted a decline in insect abundance and diversity leading to possible disastrous consequences on the food chain and ecosystems functioning. The need for monitoring technologies to reinforce observations is growing. In this thesis, a non-invasive, open-source and automated solution for insect imaging has been developed. The imaging process takes pictures from 3 different views (top, below and side) and is automated thanks to an infrared (IR) photoreflective sensor located at the device entrance. The device was designed to interface a malaise trap but is modular and could be extended to other insect traps. The power consumption of the device is relatively low ensuring an operating time ranging from 8.12 to 17.56 [h] depending on the number of insect visits. The minimum time between two specimen observations is ≈ 8.4 [s] which corresponds to a decent rate. In addition, a selection code was implemented for post-processing the set of images captured after the IR sensor trigger. This code (WMSE2) aims at keeping only the image where the insect is the most visible. 100% of correct selection were observed under luminance and noise perturbations (same on all images from the set) ranging from 0 to 80 which is the level of variation measured in real conditions operation. The accuracy drops to ≈ 80% when subjected to higher perturbations.