Road extraction from remote sensing images is significant for urban planning, intelligent transportation, and vehicle navigation. However, it is challenging to automatically extract roads from remote sensing images because the scale difference of roads in remote sensing images varies greatly, and slender roads are difficult to identify. Moreover, the road in the image is often blocked by the shadows of trees and buildings, which results in discontinuous and incomplete extraction results. To solve the above problems, this paper proposes a multiscale feature encoding and long-range context-aware network (MECA-Net) for road extraction. MECA-Net adopts an encoder–decoder structure and contains two core modules. One is the multiscale feature encoding module, which aggregates multiscale road features to improve the recognition ability of slender roads. The other is the long-range context-aware module, which consists of the channel attention module and the strip pooling module, and is used to obtain sufficient long-range context information from the channel dimension and spatial dimension to alleviate road occlusion. Experimental results on the open DeepGlobe road dataset and Massachusetts road dataset indicate that the proposed MECA-Net outperforms the other eight mainstream networks, which verifies the effectiveness of the proposed method.

The leakage of hazardous chemicals and toxic volatile substances in the atmosphere may cause serious consequences such as explosion and poisoning. To identify the unknown leakage locations and gas compositions, a mobile robot system to trace the leak source in the outdoor was investigated. First, two bionic searching algorithms, Zigzag and Silkworm algorithms, were tested with outdoor experiments for locating the leak source. The results showed that the locating errors of these two algorithms were within 0.5 m in 10 by 20 m search space, but the failing ratio of Zigzag and Silkworm algorithm was still high (about 40–50%). Therefore, an improved tracing algorithm combining the Silkworm and Zigzag algorithm, called as zigzag–Silkworm algorithm, was proposed. Compared with Silkworm and Zigzag algorithms, zigzag–Silkworm algorithm had a higher success ratio of 80% in outdoor source tracing tests, and the searching efficiency was enhanced, the efficiency parameter L: L₀ has improved from 2.58 for Silkworm and 2.66 for Zigzag to 2.17 for zigzag–Silkworm. Then, in order to identify the composition of the leaked gases during the source tracing, an artificial olfaction system (AOS) based on the gas sensor array and support vector machine was set on the mobile robot. The test results in the source tracing experiments with ammonia and ethanol emissions indicated that the recognition accuracy of emission gases reached to 99% with AOS equipped on the robot. Therefore, the mobile robot system equipped with the zigzag–Silkworm algorithm and the AOS is feasible to trace the leakage source and identify the emission composition in the outdoor leakage event with good performance in efficiency and accuracy although some underlying problems still need to be addressed in future work.

The absolute radiometric accuracy of earth-observing camera is crucial for the applications of natural resources, environment, agriculture and other industries. To continue the progress in this filed, a lunar surface reflectance based radiometric calibration approach is given in this paper. We chose IIM, M3, SP lunar models as references and GF-4 VNIR camera as sensor under calibrating. The lunar calibration sites were MS-2 site, Apollo-16 site and CE-3 site. The equivalent reflectance models of lunar were retrieved by Multiplying and integrating with the spectral response function of VNIR camera. Absolute radiometric calibrations with the equivalent reflectance models of lunar were carried out for 520–590nm, 630–690nm and 770–890nm spectral bands. The ground-based validation experiments were conducted with low, medium and high reflectance targets. The calibration accuracy was evaluated by comparing the relative errors of derived radiance after radiometric calibration with the benchmarks of TOA radiance transferred by in-situ measured reflectance. Lunar-based calibration models, lab and on-orbit filed-based models were used to compare the relative errors between proposed method and traditional way. The results showed that using IIM lunar model had better radiometric accuracy than other models, and SP model had the similar performance with traditional on-orbit filed-based model. The results indicated that using lunar to calibrate the earth-observing camera had the capability to improve the radiometric calibration accuracy.

Keywords: GF-4 camera, Lunar radiation calibration Starting from the lunar observation requirements of the GF-4 satellite, the main index such as the resolution, the imaging field, the reflect radiance and the imaging integration time are analyzed combined with the imaging features and parameters of this camera. The analysis results show that the lunar observation of GF-4 satellite has high resolution, wide field which can image the whole moon, the radiance of the pupil which is reflected by the moon is within the dynamic range of the camera, and the lunar image quality can be guaranteed better by setting up a reasonable integration time. At the same time, the radiation transmission model of the lunar radiation calibration is trace and the radiation degree is evaluated.