Search

In order to improve the detection accuracy of surrounding targets by self-driving vehicles, a target object detection algorithm (monocular-bird's eye view fusion, Mono-BEVFusion) with fusion of lidar and monocular image data on bird's eye view features is proposed. To construct camera BEV features, the algorithm builds a simple and efficient depth prediction network to predict the depth of camera features, which is supervised with depth truth values based on an explicit supervision method. When constructing the lidar BEV features, the laser point cloud is voxelized into a columnar grid and transformed into BEV features, and the BEV feature fusion network is designed to fuse the laser point cloud BEV features with the camera BEV features, and the fused features are inputted into the target detection framework to obtain the detection results of the target objects (cars, pedestrians, and cyclists). The Mono-BEVFusion algorithm is evaluated using KITTI dataset and real-vehicle road-harvesting data, and the experimental results show that the algorithm improves the mean average precision by 2.90 percentage points compared with existing fusion algorithms, in which the single detection precision of the car class and the pedestrian class is improved by 3.38 percentage points and 4.13 percentage points, respectively. The Mono-BEVFusion algorithm has more stable detection effect on the occluded targets or the targets with a long distance, and can effectively avoid the leakage detection phenomenon of single sensor, which has better practical application value. [ABSTRACT FROM AUTHOR]

In order to solve the problem of obtaining 3D terrain data of shoal and tidal flat in the large area of filling project, through the research and development of high -precision altimetry technology combining UAV aerial photography with laser radar and unmanned ship bathymetry, the laser point cloud data was used in the phase control of photogrammetry, the processing flow and technology of LIDAR data were improved, the intensity of work was reduced, the work efficiency and precision were improved. The unmanned ship survey system was autonomous integrated to adapt to the filling engineering, which can realize the bathymetric survey in the blind area by traditional survey ships. The practical application shows that the unmanned measurement technology formed by the combination of UAV and unmanned ship measurement technology can achieve high-efficient 3D data acquisition, and the precision can meet the requirements of the topographic survey specification. [ABSTRACT FROM AUTHOR]

Aiming at the problems of low accuracy and poor robustness of existing LiDAR odometry methods in outdoor dynamic road scene, this paper proposed a scheme of odometry and mapping based on 3D LiDAR and MEMS IMU. The scheme was consist of four sequential module, namely data preprocessing module,feature extraction considering dynamic obstacles module, laser odometry module and laser mapping module. By using the KITTI data and real scene data, it carried out the qualitative and quantitative accuracy evaluation of the improved scheme and the current popular LOAM and Lego-LOAM schemes. The test results show that the accuracy of the improved scheme is better than LOAM scheme under the real-time requirements of the system, which is equivalent to Lego-LOAM scheme. [ABSTRACT FROM AUTHOR]

：Objectives: The dangerous rock masses on the high steep slope are easy to lose stability, which threatens the safety of water conservancy projects. Therefore, the rapid, accurate and convenient identification of dangerous rock mass on the high steep slope is significant to the construction of hydropower projects. As a new equipment of surface observation, unmanned aerial vehicle (UAV) can be equipped with laser radar （LiDAR） to obtain high ‐ density and high ‐ precision point cloud data of large areas on high steep slopes. Methods: A radar laser scan of the high steep slope near Lianghekou Hydropower station is carried out by using UAV, and three areas are scanned, including one on the left bank and two on the right bank. Firstly, on the basis of the ground points extracted from the point cloud data were decomposed and smoothed, and then the edge points are obtained through geometric features. After that, the point cloud is clustered based on the edge points to form the candidate object. Finally, we generate the digital elevation model, by which we can calculate the spatial feature of the object to extract the dangerous rock mass. Results: In the three areas explored in this research, a total of 47 potentially dangerous rock masses are detected. The above identification method is used to detect dangerous rock mass, and 16 dangerous rock masses are found. 5 dangerous rock masses are detected in scanning areas 1 and 2, and 6 dangerous rock masses are detected in scanning area 3. Conclusions: UAV equipped with LiDAR can obtain the spatial coordinate data of large area on high steep slope accurately, and thus the spatial information of isolated dangerous rock mass can be obtained.Combined with the area, maximum height difference and volume, dangerous rock mass can be extracted quickly and effectively. Engineering example shows that our proposed method has strong applicability and reliability; and it can be used to quickly identify isolated dangerous rock mass on high steep slope. [ABSTRACT FROM AUTHOR]

The fruit industry would suffer a great shock, due mainly to the fact that its yield relies heavily on high labor inputs, but the rural population is aging with the ever-increasing development of cities in China. Autonomous production can bring an effective solution to such issues, further promoting the precision management in orchards. 3D light detection and ranging (LiDAR) sensor has made a much greater contribution to the autonomous navigation in the information acquisition for orchards, compared with the traditional 2D laser scanner. Specifically, LiDAR is a commonly-used remote sensing technique, where a laser is used to measure the distance to an illuminated target. In this study, an inter-row robot navigation was thus proposed in an orchard using 3D LiDAR. The complex three-dimensional scene was treated effectively, particularly with the dense canopy and trunks occluded by branches. A 3D LiDAR detection device was used to collect the environment information at first, and a pass-through filter was then used to correct the region of interest, where the noise was removed from the positioning task. Euclidean clustering was used to recognize the fruit trees around the robot, assuming that the tree branches were subjected to the normal distribution in the vertical direction. Body centers of trees were equivalent to the position of trees. Random sampling consensus and the least square were selected to fit the tree data using the parallelism between the tree rows. A complementary fusion was also put forward to combine two fittings. The centerline between tree rows was calculated and then treated as the target navigation line. In addition, a pure pursuit algorithm was refined using the differential chassis, considering the looking-forward distance and heading deviation. The validation experiments were carried out in a simulated hedgerow orchard and a real pear orchard. It was found that the tree rows successfully fitted with great ability to resist the interference from the environment in the first scenery. The heading positioning deviation was within 1.65°, and the lateral deviation was within 6.1 cm, when the robot walked along the centerline at a speed of 0.33 m/s. The tracking system automatically followed the centerline with a speed of 0.43 m/s, with an absolute lateral deviation of within 15cm. In the second scenery, the tracking system followed the centerline with two speeds of 0.68 and 1.35 m/s, where the absolute lateral deviations were not beyond 21.3 cm and 22.1 cm, respectively. The tracking system can be expected to serve as the automatic navigation with good robustness in standard orchards, including the hedgerow or complex three-dimensional orchard. The fruit industry would suffer a great shock, due mainly to the fact that its yield relies heavily on high labor inputs, but the rural population is aging with the ever-increasing development of cities in China. Autonomous production can bring an effective solution to such issues, further promoting the precision management in orchards. 3D light detection and ranging (LiDAR) sensor has made a much greater contribution to the autonomous navigation in the information acquisition for orchards, compared with the traditional 2D laser scanner. Specifically, LiDAR is a commonly-used remote sensing technique, where a laser is used to measure the distance to an illuminated target. In this study, an inter-row robot navigation was thus proposed in an orchard using 3D LiDAR. The complex three-dimensional scene was treated effectively, particularly with the dense canopy and trunks occluded by branches. A 3D LiDAR detection device was used to collect the environment information at first, and a pass-through filter was then used to correct the region of interest, where the noise was removed from the positioning task. Euclidean clustering was used to recognize the fruit trees around the robot, assuming that the tree branches were subjected to the normal distribution in the vertical direction. Body centers of trees were equivalent to the position of trees. Random sampling consensus and the least square were selected to fit the tree data using the parallelism between the tree rows. A complementary fusion was also put forward to combine two fittings. The centerline between tree rows was calculated and then treated as the target navigation line. In addition, a pure pursuit algorithm was refined using the differential chassis, considering the looking-forward distance and heading deviation. The validation experiments were carried out in a simulated hedgerow orchard and a real pear orchard. It was found that the tree rows successfully fitted with great ability to resist the interference from the environment in the first scenery. The heading positioning deviation was within 1.65°, and the lateral deviation was within 6.1 cm, when the robot walked along the centerline at a speed of 0.33 m/s. The tracking system automatically followed the centerline with a speed of 0.43 m/s, with an absolute lateral deviation of within 15cm. In the second scenery, the tracking system followed the centerline with two speeds of 0.68 and 1.35 m/s, where the absolute lateral deviations were not beyond 21.3 cm and 22.1 cm, respectively. The tracking system can be expected to serve as the automatic navigation with good robustness in standard orchards, including the hedgerow or complex three-dimensional orchard. [ABSTRACT FROM AUTHOR]

We explore the method for object-oriented seasonal saltwater lake wetland natural boundary location using multi-source data including high-resolution remote sensing images, DEM(digital elevation model) and so on,taking Yanchi County of Ningxia Province as an example. Taking high-resolution remote sensing images,high spatial resolution DEM generated by LiDAR(light detection and ranging) data, general investigation data on geography and national status and the second national wetland resources investigation data as data sources, we construct the decision rule model with the LBV transform features, NDVI(normalized difference vegetation index), DEM, slope and other features and execute to get the seasonal saltwater lake wetland􀆳s natural boundary. Then we analyze the impact of DEM in different spatial resolutions in seasonal saltwater lake wetland boundary extraction. Experimental results show that DEM􀆳s participation in appreciate high spatial resolution in seasonal saltwater lake wetland natural boundary extraction can improve the accuracy of boundary location. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Heilongjiang University of Science & Technology is the property of Journal of Heilongjiang University of Science & Technology Editorial Department and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

To the question that multisource features' contribution to classification is not explicit in airborne LiDAR system data, based on object oriented data mining, this paper proposed a method to select features for classification using Random Forest. It's proved that the features' contribution can be evaluated correctly and the selected features can still make a high classification accuracy. [ABSTRACT FROM AUTHOR]

Based on the characteristics of the morphological filter and the TIN-based progressive filter, a high-quality LiDAR point cloud filtering algorithm combining Morphological grayscale reconstruction and TIN Models is proposed in this paper. Its main strategies are: 1 Implementing morphological grayscale reconstruction with a priority of Type I Error and non-minimum suppression. In this step, LiDAR point clouds are tagged as Reliable terrain points G, suspicious terrain points S and suspicious Non-terrain points NG; 2 Suspicious non-terrain points are further tagged based on the iterative order of Morphological grayscale reconstruction. In this step, small and constant height interval is used to filter the possible non-terrain points at different elevation; 3 Constructing the initial TIN from points G and further filtering points S and NG points, respectively, by adaptively adjusting the parameters of the ground point criterion at associated point layer. We did an experiment with 15 ISPRS test data sets and assessed the results with the standard criterion as found in the literature. The result shows that proposed filtering algorithm dramatically improved filtering quality, even for complex terrain. [ABSTRACT FROM AUTHOR]

Two types of approach called region-growing and random sample consensus have been proposed for automatic building roof extraction. They both, however, have drawbacks. In this paper, an hybridized method is proposed to take advantage of both algorithms7 strengths so that building roofs can be extracted more precisely and efficiently. First, we calculate the normal and rough features from LiDAR data for the coarse extraction of building roofs. Second, precise roof-extraction is performed using an extended RANSAC method which takes the coarse extraction results as the priori-knowledge, and integrates a region growing method. Finally, a poll strategy is adopted to solve the competition problem. The experimental results show that our method can extract intact building roofs in a highly automated manner. [ABSTRACT FROM AUTHOR]