Multi-nighttime-light data comparison analysis based on image quality values and lit fishing vessel identification effect

Abstract

Fisheries provide high-quality protein for many people, and their sustainable use is of global concern. Light trapping is a widely used fishing method that takes advantage of the phototropism of fish. Remote sensing technology allows for the monitoring of lit fishing vessels at sea from the air at night, which supports the sustainable management of fisheries. To investigate the potential of different nighttime light remote sensing data for lit fishing vessel identification and applications, we used the fuzzy evaluation method to quantitatively assess images in terms of their radiometric and geometric quality, and Otsu’s method to compare the effects of lit fishing vessel identification. Three kinds of nighttime lighting data from the Defense Meteorological Satellite Program/Operational Linescan System (DMSP/OLS), Visible infrared imaging radiometer suite day/night band (VIIRS/DNB), and Luojia1-01(LJ1-01) were analyzed, compared, and application pointers were constructed. The results are as follows. ①In the image radiation quality evaluation, the information entropy, clarity, and noise performance of the LJ1-01 image are higher than those of the DMSP/OLS and VIIRS/DNB images, where the information entropy value of the LJ1-01 image is nearly 10 times that of VIIRS/DNB and 23 times that of DMSP/OLS. The average gradient value is 14 times that of the image from VIIRS/DNB and 1,600 times that of DMSP/OLS, while its noise is only about 2/3 of the VIIRS/DNB image and 1/3 of the DMSP/OLS image. In the geometric quality assessment, the geometric positioning accuracy and ground sampling accuracy of the VIIRS/DNB image is the best among the three images, with a relative difference percentage of 100.1%, and the LJ1-01 and DMSP/OLS images are relatively lower, at 96.9% and 92.3%, respectively. ② The detection of squid fishing vessels in the Northwest Pacific is taken as an example to compare the identification effects of three types of data: DMSP/OLS, VIIRS/DNB, and LJ1-01. Among these data, DMSP/OLS can effectively identify the position of the lit fishing boat, and VIIRS/DNB images can accurately estimate the spatial position and number of lit fishing boats with large distances. However, in the case of fishing boats gathering or clustering, the number of fishing vessels could not be identified. This led to the detected number of lit fishing vessels being less than the real value. For the VIIRS/DNB and LJ1-01 images with a 5′×8′ span in the same spatiotemporal range using the same batch of pelagic squid fishing vessels, LJ1-01 extracted 18 fishing vessels. VIIRS/DNB extracted 15, indicating that LJ1-01 can distinguish multiple fishing vessels in the lighted overlapping area, thus accurately identifying the number of fishing vessels. The application pointing table generated based on the results of the three data analyses can provide a reference for sensor/image selection for nighttime light remote sensing fishery applications and a basis for more refined fishing vessel identification, extraction, and monitoring.