A set of customized identification process of coastal and marine microplastics garbage using UAV and hyperspectral camera is proposed. And show a series of results obtained from field tests on the coast of Sassari, Italy. In the experiment, DJI-M600 UAV is equipped with a 900- 1700 nm linear push-broom hyperspectral imager (self-made, using NIR 17 spectrometer of SpecIM+Bobcat 32 SWIR camera of +Xenics, with resolution of 256 * 320 and maximum frame rate of 50Hz) to collect data. Subsequent image stitching, geographic registration, orthographic image production and other processing are all carried out in the office.
The generation of hyperspectral image cube is realized by stitching synchronously collected visible light images, and the generation of images is realized by performing the same translation, rotation and scaling operations on hyperspectral data through correlation-based registration.
The identification of plastic waste can be realized by selecting statistically relevant features, using LDA analysis and training through manually selected samples.
In this paper, experiments have been carried out on beaches and nearby waters. The experimental results show that the classification method in this paper can effectively identify and distinguish PE (polyethylene) and PET (polyethylene terephthalate) items.
In the future, I hope to realize real-time processing of this method, and get the results by online processing and analysis on the machine, which is convenient for rapid results in UAV monitoring.
The DJI-M600 UAV is equipped with a self-made hyperspectral imager, an Intel NUC computer and a matching visible light camera, with a total weight of 2.5kg and a power of 100W, as shown in the following figure.
Hyperspectral imager: 900- 1700 nm, self-made, using NIR 17 spectrometer of+SPECIM of +Xenics Company and Bobcat 32 Swir camera, with resolution of 256 * 320, maximum frame rate of 50Hz and dynamic range of 8 bits.
The following figure shows the spectral curves of PE and PET, and it can be seen that there is still a big difference between them.
The picture below shows the results of plastic waste detection and classification on grassland.
The picture below shows the detection results of bare plastic waste.
The picture below shows the plastic test results in the sea near the beach.
The picture below shows the test results at Platamona Beach.
The picture below shows the test results at Porto Ferro Beach.