16:30 〜 16:45
[HCG19-04] Detection and characterisation of plastic pollution on continental surfaces
キーワード：Hyperspectral, UAV, Plastic pollution, Continental surfaces, Detection, Classification
The annual production of plastic has been steadily increasing since the 1960s, coinciding with the rise in plastic pollution. Given that 80% of the plastics observed in the oceans originate from the continents [1], there is a pressing need to map and collect these waste materials on land before they reach the oceans. However, while marine pollution is widely documented, few studies are devoted to mapping plastic waste on continental surfaces using optical remote sensing methods. In particular, hyperspectral imaging (HSI) is a technique that combines spectroscopy and digital imaging, providing information with high spectral and spatial resolution. Each pixel is associated with an electromagnetic spectrum, allowing the characterization of the materials composing the pixel by analyzing their unique spectral response. The choice of HSI is based on the diversity of plastic sizes (microplastics of size less than 5 mm and macroplastics of size greater than 5 mm) and the variety of plastic types (PE, PET, PP, PS, etc.), which have seemingly similar spectra but differ when the spectral resolution is sufficiently fine. Previous works show that HSI enables effective detection of plastic waste on various soil types [2, 3]. Combined with UAV technology, HSI would allow fast and efficient surveys of large areas contaminated by plastic waste including coastal, urban and vegetation environment.
Yet, to our knowledge, studies conducted with HSI regarding plastic pollution on continental surfaces have been carried out almost exclusively in the laboratory under controlled conditions or airborne by aircraft. Therefore, the challenge is to evaluate the potential of UAV coupled with HSI for mapping plastic waste in various contexts. To this end, three hyperspectral datasets were acquired: one in the laboratory with a spatial resolution of 1.3 mm, and two from drone campaigns over a coastal area and a vegetated scene with spatial resolutions of a few centimeters. Several detection and identification methods were developed (spectral indices, spectral distances, machine learning, and spectral unmixing) using the database constructed in the laboratory as well as spectral reflectance data from the literature [4]. This paper focuses on the drone campaign in a coastal environment. Three plastic workshops with samples of different sizes (ranging from 1 to 50 cm) and types (PE, PET, PP, PS, etc.) were set up on the sand and imaged at various altitudes with spatial resolutions between 3 and 10 cm in the 1-1.7 µm range [Fig. 1a]. The work is divided into three phases: (i) detection of all plastic samples placed on the scene; (ii) identification of the types of detected plastics; (iii) evaluation of the performance of the different methodological approaches using ground truth references. An initial example of a detection map using the spectral indices method is shown in Fig. 1b. In this case, the sPA (detection performance metric ranging from 0.5 to 1) is 0.87, indicating a low false alarm rate. This preliminary result demonstrates the potential of UAV technology combined with HSI in this application context.
References:
[1] Institut Heinrich Böll Stiftung. « Atlas du Plastique | Heinrich Böll Stiftung | Bureau Paris - France », https://fr.boell.org/fr/atlas-du-plastique
[2] Shan, Jiajia et al. « A novel way to rapidly monitor microplastics in soil by hyperspectral imaging technology and chemometrics ». https://doi.org/10.1016/j.envpol.2018.03.026
[3] Ali, Mansurat A. et al. « Critical evaluation of hyperspectral imaging technology for detection and quantification of microplastics in soil ». https://doi.org/10.1016/j.jhazmat.2024.135041
[4] Garaba, Shungudzemwoyo P. et al. « An airborne remote sensing case study of synthetic hydrocarbon detection using short wave infrared absorption features identified from marine-harvested macro- and microplastics ». https://doi.org/10.1016/j.rse.2017.11.023
Yet, to our knowledge, studies conducted with HSI regarding plastic pollution on continental surfaces have been carried out almost exclusively in the laboratory under controlled conditions or airborne by aircraft. Therefore, the challenge is to evaluate the potential of UAV coupled with HSI for mapping plastic waste in various contexts. To this end, three hyperspectral datasets were acquired: one in the laboratory with a spatial resolution of 1.3 mm, and two from drone campaigns over a coastal area and a vegetated scene with spatial resolutions of a few centimeters. Several detection and identification methods were developed (spectral indices, spectral distances, machine learning, and spectral unmixing) using the database constructed in the laboratory as well as spectral reflectance data from the literature [4]. This paper focuses on the drone campaign in a coastal environment. Three plastic workshops with samples of different sizes (ranging from 1 to 50 cm) and types (PE, PET, PP, PS, etc.) were set up on the sand and imaged at various altitudes with spatial resolutions between 3 and 10 cm in the 1-1.7 µm range [Fig. 1a]. The work is divided into three phases: (i) detection of all plastic samples placed on the scene; (ii) identification of the types of detected plastics; (iii) evaluation of the performance of the different methodological approaches using ground truth references. An initial example of a detection map using the spectral indices method is shown in Fig. 1b. In this case, the sPA (detection performance metric ranging from 0.5 to 1) is 0.87, indicating a low false alarm rate. This preliminary result demonstrates the potential of UAV technology combined with HSI in this application context.
References:
[1] Institut Heinrich Böll Stiftung. « Atlas du Plastique | Heinrich Böll Stiftung | Bureau Paris - France », https://fr.boell.org/fr/atlas-du-plastique
[2] Shan, Jiajia et al. « A novel way to rapidly monitor microplastics in soil by hyperspectral imaging technology and chemometrics ». https://doi.org/10.1016/j.envpol.2018.03.026
[3] Ali, Mansurat A. et al. « Critical evaluation of hyperspectral imaging technology for detection and quantification of microplastics in soil ». https://doi.org/10.1016/j.jhazmat.2024.135041
[4] Garaba, Shungudzemwoyo P. et al. « An airborne remote sensing case study of synthetic hydrocarbon detection using short wave infrared absorption features identified from marine-harvested macro- and microplastics ». https://doi.org/10.1016/j.rse.2017.11.023