Detection of Fishing Boats in the Northwest Pacific Using Satellite Nighttime Image

Transcription

1 2018 International Conference on Communication, Network and Artificial Intelligence (CNAI 2018) ISBN: Detection of Fishing Boats in the Northwest Pacific Using Satellite Nighttime Image Hao TIAN 1, Yang LIU 1,*, Yong-Jun TIAN 1, Guan-yu CHEN 1, Jian-chao LI 1, Shigang LIU 1, Lu-xin YAN 1, Yuan LI 2 and Long-Shan LIN 2 1 College of Fisheries, Ocean University of China, Qingdao , China 2 Third Institute of Oceanography, state Oceanic Administration, Xiamen , China *Corresponding author Keywords: Pacific saury, Fishing boat identification, Remote sensing, VIIRS/DNB. Abstract. In recent years, Spatial and temporal dynamic monitoring of fishing boats has become an important data source for understanding the distributional dynamics of fisheries and for combating illegal fishing. Visible infrared imaging radiometer suite (VIIRS) Day / night band (DNB) nighttime remote sensing images can be used to monitor night fishing lights. In the study that we are reporting here, spike detection and threshold segmentation techniques were used to identify Pacific saury fishing boats employing night lights as fish attractors. GIS tools were then used to extract and analyze the fishing boat position and operation status. The results show that the method proposed in this study can effectively identify the location and operation status of Pacific saury fishing boats, and provide useful information for further understanding the spatial distribution and dynamics of Pacific saury fishing boats in the northwest Pacific region. Introduction Pacific saury (Cololabis saira) is one of the most important commercial pelagic fish species being harvested in the northwestern Pacific Ocean region [1]. Saury maintain a depth of about 15m below the sea surface during the daytime and ascend to the sea surface at night. Pacific saury fishing boats attract and concentrate Pacific saury with strong lights, then capturing them with pole-mounted dip nets. Pacific saury represent an economically important resouce for Chinese, Japanese, Russian and Korean fisheries operating in the traditional saury fishing grounds northeast of Japan [2]. In 2014, the catch of saury was 620,000 tons (statistics data from FAO). As early as the 1970s, some scholars discovered that low light imaging detectors can serve to monitor light-fishing boats [3]. The first system to globally monitor night lights was the defense meteorological satellite program's operational linescan system (DMSP/OLS) [4]. Due to its low resolution, fewer bands, and lack of sensor radiance calibration [5], DMSP/OLS data can only be used for qualitative research and cannot be used for quantitative detection of the number of nighttime light-fishing boats. The US National Polar Orbiting Partnership (NPP) satellite was launched on October 28, 2011, carrying a visible infrared imaging radiometer suite (VIIRS). This VIIRS is a 22- band visible/infrared sensor that offers a wider swath width (3,000 versus 2,330 km) [6]. NPP crosses the equator at about 0130 and 1330 local time, with VIIRS covering the entire Earth twice daily [7]. VIIRS also has a day/night band (DNB) that is sensitive to very low levels of nighttime visible light such as may be produced by moonlight reflecting off low clouds, fog, dust, ash plumes, and snow cover. In addition, VIIRS detects light emissions from cities, ships, oil flares, and lightning flashes [8]. 65

2 Data and Methods Satellite Data Sensor data records (SDRs) used in this study are calibrated image files that provide the radiance information of DNB. The spectral range of the DNB waveband is between 400 and 900 nm, which allows it to collect the type of light coming from fishing boats [9]. The spatial resolution of DNB data is 742m, and the original radiance unit is W/ (sr*cm -2 ) [10]. All the satellite data was downloaded from the National Oceanic and Atmospheric Administration (NOAA), The DNB data was opened using ENVI 5.3. The Pacific saury fishing boats were detected and extracted by ArcGIS Algorithm Flow Data Preprocessing. The data preprocessing includes geographic lookup table (GLT) geometric correction, data format conversion and contrast adjustment. The GLT geometric correction is the first step to make each pixel of the original DNB data correspond to its latitude and longitude coordinates. Then the DNB data format is converted into dat so that it can be opened by ArcGIS. The cloud will not reflect the moon radiance and it will be difficult to detect when the moonlight is weak. So the third step is adjusting the contrast of the DNB image to asertain whether or not there is a cloud in the research area. Data Selection. Because clouds and moonlight will affect the identification of saury fishing boats [11], we selected the cloudless and new moon nighttime data in study area. This paper uses the DNB images of October 11th, 2016 at 00:27:56 as an example to illustrate the detection method of the saury fishing boats (Figure 1). Figure 1. DNB IMAGE. Removing Lightning. The lightning in the VIIRS DNB data is a horizontal light stripe with a width of more than 16 lines, usually more than 24 pixels long. Lightning will interfere with the detection of a night-running saury fishing vessel. So removing the lightning in the VIIRS DNB images to improve the detection accuracy of the fishing boat is necessary. 66

3 Calculation the Spike Median Index (SMI). In this study, the spike median index (SMI) proposed by Elvidge et al.(2015) was used to enhance the difference between the lighting pixels radiance value and the background pixels radiance value, thereby facilitating the detection and extraction of saury night-light fishing boat information. The SMI image is obtained by subtracting the amplified image from the median-filtered image. Gray Image Conversion. In actual fishing of Pacific saury in the Northwestern Pacific Ocean, the saury fishing boats usually have two states at night: fishing or moving. In this study, gray image conversion was performed to discriminate the two operating conditions of the Pacific saury fishing boats based on the brightness of the pixel. Firstly, we use log10 to increase the contrast of DNB images and highlight light pixels [12]. According to the study of Cozzolion et al. (2016), the radiance values of the pixels after log10 can converted to gray values (0 to 255) Threshold Selection. Selecting an appropriate threshold value makes it possible to distinguish between the fishing boats and the background pixels. According to the selection, SMI=0.7 and SMI=9 are the thresholds we selecte. Segmentation. Segmentation is a process that can classify the pixels of image by different criteria [13]. The purpose of this study was to separate the saury light fishing boats from background pixels, so we segmented the pixels whose values threshold. Reclassify. Because of the different operating status of the saury fishing boats, the fishing boats detection was divided into two levels: strong detection represents the fishing boats that are catching saury; weak detection represents the fishing boats are moving to find saury. Pixels whose SMI 0.7 are defined as weak detection while SMI 9 are defined as strong detection. Vector File Extraction. Overlarge volume of DNB images has caused great difficulties for data storage and use. In this study, the layer which contains the position information of the fishing vessel was stored as a shp file. Results Gray Image Results The gray image conversion is shown in figure 2 and the brightness of lighting pixels is clearly divided into 2 levels: the brighter pixels may be the boats that are fishing and the darker pixels may be the moving boats. Figure 2. Gray image results, the light pixels are divided in two brightness, the brighter pixels are fishing boats and the darker pixels are moving boats. 67

4 Fishing Boats Detection Results According to the different thresholds selected from quality grading, identified Pacific saury fishing boats were extracted from DNB images. The results are shown in figure 3. At the time of 00:27:56 on October 11, 2016, the region we selected contains 77 strong detection points and 124 weak detection points. Figure 3. Fishing boats detection results, 77 fishing boats and 124 moving boats have been extracted. Discussion Our study has found that the VIIRS DNB images, while affected by cloud and moonlight radiance, are nevertheless able to effectively identify nighttime light fishing boats under conditions of sufficiently low moonlight intensity and low level of cloud cover. Although a cloud mask is able effectively remove clouds, the information about the fishing boats blocked by those clouds will be removed at the same time. At the full moon, moonlight reflected by clouds will also adversely affect the detection of fishing boats [14]. However, Yamaguchi et al. (2016) found that the brightness and temperature information contained in a DNB image and a short wave infrared image whose wavelength is 3.7 μm can reduce the influence of partial cloud cover [15]. In monitoring Pacific saury fishing activity, it was found that the fishing boats had two distinguishable kinds of operating conditions at night: fishing or moving. The two operating states with different light emission brightness were confirmed in the gray image. This study finds and verifies the distinction thresholds for these two operating states: SMI=0.7 is the threshold for distinguishing between the marine background and moving state fishing boats; SMI=9 is the distinguishing threshold for moving and fishing state of Pacific saury fishing boats. The method proposed in this study was applied to DNB data in the Northwestern Pacific Ocean on October 11, 2016 to detect the Pacific saury fishing boats at 00:27:56. The result is that in the region we selected, the number of Pacific saury fishing boats in the fishing state was 77 and moving state was 124. Strong fishing light in the DNB image will illuminate the surrounding area and form false detections. Therefore, the number of weak detection in this study is large. The misdetection of Pacific saury fishing boats caused by halos can be effectively eliminated combined with the vessel monitoring system (VMS). 68

5 Conclusion Combined with the distribution range of the fish species and the fishing conditions, the DNB images can be used to analyze the types of night-light fishing boats. There are two types of working status for the Pacific saury fishing boats at night: fishing state and moving state. The two statuses exhibit different brightness levels in the DNB image, and the brightness of the fishing state is higher. Acknowledgment Supported by the Fundamental Research Funds for the Central Universities ( ); China Postdoctoral Science Foundation (187202); And the National Program on Global Change and Air-Sea Interaction (GASI-02-PAC-YDspr/sum/aut) References [1] Y. Tian, T. Akamine, and M. Suda. "Variations in the abundance of Pacific saury ( Cololabis saira ) from the northwestern Pacific in relation to oceanic-climate changes." Fisheries Research 60.2(2003): [2] C.T. Tseng, et al. "Sea surface temperature fronts affect distribution of Pacific saury( Cololabis saira ) in the Northwestern Pacific Ocean." Deep-Sea Research Part II 107(2014): [3] T.A. Croft. "Nighttime Images of the Earth from Space." Scientific American 239.1(1978): [4] C.D. Elvidge, K.E. Baugh, M. Zhizhin, and F.C. Hsu. "Why VIIRS data are superior to DMSP for mapping nighttime lights." Proceedings of the Asia-Pacific Advanced Network 35(2013): [5] Y. Liu, S.I. Saitoh, T. Hirawake, H. Igarashi, and Y. Ishikawa. "Detection of Squid and Pacific Saury fishing vessels around Japan using VIIRS Day/Night Band image." Asia pacific Advanced Network 2015: [6] S.D. Miller, et al. "Illuminating the Capabilities of the Suomi National Polar-Orbiting Partnership (NPP) Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band." Remote Sensing 5.12(2013): [7] E. Cozzolino, and C.A. Lasta. "Use of VIIRS DNB satellite images to detect jigger ships involved in the Illexargentinus, fishery." Remote Sensing Applications Society & Environment 4(2016): [8] G. Guo, et al. "Identification for operating pelagic light-fishing vessels based on NPP/VIIRS low light imaging data." Transactions of the Chinese Society of Agricultural Engineering 33.10(2017): (In Chinese with English abstract). [9] K. Shi, et al. "Evaluating the Ability of NPP-VIIRS Nighttime Light Data to Estimate the Gross Domestic Product and the Electric Power Consumption of China at Multiple Scales: A Comparison with DMSP-OLS Data." Remote Sensing 6.2(2014): [10] C. Cao, et al. "Suomi NPP VIIRS sensor data record verification, validation, and long-term performance monitoring." Journal of Geophysical Research Atmospheres (2013): 11-11,678. [11] K. Baugh, C.D. Elvidge, T. Ghosh, and D. Ziskin. "Development of a 2009 stable lights product using DMSP-OLS data." Proceedings of the Asia-Pacific Advanced Network 30(2010). [12] H.M. Wechsler. "Digital image processing, 2nd ed." Proceedings of the IEEE 69.9(2005):

6 [13] G.N. Srinivasan, and G. Shobha. Segmentation techniques for target recognition.world Scientific and Engineering Academy and Society (WSEAS), [14] T.J. Kopp, et al. "The VIIRS Cloud Mask: Progress in the first year of S-NPP toward a common cloud detection scheme." Journal of Geophysical Research Atmospheres 119.5(2014): [15] T. Yamaguchi, I. Asanuma, J.G. Park, K.J. Mackin, and J. Mittleman. "Estimation of vessel traffic density from Suomi NPP VIIRS day/night band." OceansIEEE, 2016: