Airborne Light Detection And Ranging (LiDAR) systems usually operate at a monochromatic wavelength measuring the range and the strength of the reflected energy (intensity) from objects. Recently, multispectral LiDAR sensors, which acquire data at different wavelengths, have emerged. This allows for recording of a diversity of spectral reflectance from objects. In this context, we aim to investigate the use of multispectral LiDAR data in land cover classification using two different techniques. The first is image-based classification, where intensity and height images are created from LiDAR points and then a maximum likelihood classifier is applied. The second is point-based classification, where ground filtering and Normalized Difference Vegetation Indices (NDVIs) computation are conducted. A dataset of an urban area located in Oshawa, Ontario, Canada, is classified into four classes: buildings, trees, roads and grass. An overall accuracy of up to 89.9% and 92.7% is achieved from image classification and 3D point classification, respectively. A radiometric correction model is also applied to the intensity data in order to remove the attenuation due to the system distortion and terrain height variation. The classification process is then repeated, and the results demonstrate that there are no significant improvements achieved in the overall accuracy.

The release of the world’s first dual-frequency GPS/Galileo smartphone, Xiaomi mi 8, in 2018 provides an opportunity for high-precision positioning using ultra low-cost sensors. In this research, the GNSS precise point positioning (PPP) accuracy of the Xiaomi mi 8 smartphone is tested in post-processing and real-time modes. Raw dual-frequency observations are collected over two different time windows from both of the Xiaomi mi 8 smartphone and a Trimble R9 geodetic-quality GNSS receiver using a short baseline, due to the lack of a nearby reference station to the observation site. The data sets are first processed in differential modes using Trimble business center (TBC) software in order to provide the reference positioning solution for both of the geodetic receiver and the smartphone. An in-house PPP software is then used to process the collected data in both of post-processing and real-time modes. Precise ephemeris obtained from the multi-GNSS experiment (MGEX) is used for post-processing PPP, while the new NAVCAST real-time GNSS service, Germany, is used for real-time PPP. Additionally, the real-time PPP solution is assessed in both of static and kinematic modes. It is shown that the dual-frequency GNSS smartphone is capable of achieving decimeter-level positioning accuracy, in both of post-processing and real-time PPP modes, respectively. Meter-level positioning accuracy is achieved in the kinematic mode.

Unmanned aerial vehicle (UAV) navigation has recently been the focus of many studies. The most challenging aspect of UAV navigation is maintaining accurate and reliable pose estimation. In outdoor environments, global navigation satellite systems (GNSS) are typically used for UAV localization. However, relying solely on GNSS might pose safety risks in the event of receiver malfunction or antenna installation error. In this research, an unmanned aerial system (UAS) employing the Applanix APX15 GNSS/IMU board, a Velodyne Puck LiDAR sensor, and a Sony a7R II high-resolution camera was used to collect data for the purpose of developing a multi-sensor integration system. Unfortunately, due to a malfunctioning GNSS antenna, there were numerous prolonged GNSS signal outages. As a result, the GNSS/INS processing failed after obtaining an error that exceeded 25 km. To resolve this issue and to recover the precise trajectory of the UAV, a GNSS/INS/LiDAR integrated navigation system was developed. The LiDAR data were first processed using the optimized LOAM SLAM algorithm, which yielded the position and orientation estimates. Pix4D Mapper software was then used to process the camera images in the presence of ground control points (GCPs), which resulted in the precise camera positions and orientations that served as ground truth. All sensor data were timestamped by GPS, and all datasets were sampled at 10 Hz to match those of the LiDAR scans. Two case studies were considered, namely complete GNSS outage and assistance from GNSS PPP solution. In comparison to the complete GNSS outage, the results for the second case study were significantly improved. The improvement is described in terms of RMSE reductions of approximately 51% and 78% for the horizontal and vertical directions, respectively. Additionally, the RMSE of the roll and yaw angles was reduced by 13% and 30%, respectively. However, the RMSE of the pitch angle was increased by about 13%.

Accurate and up-to-date land cover classification information is essential for many applications, such as land-use change detection, global environmental change, and forest management, among others. Unoccupied aerial systems (UASs) provide the advantage of flexible and rapid data acquisition at low cost compared to conventional platforms, such as satellite and airborne systems. UASs are often equipped with high spatial resolution cameras and/or light detection and ranging (LiDAR). However, the high spatial resolution imagery has a high information content, which makes land cover classification quite challenging. Recently, deep convolutional neural networks (DCNNs) have been effectively applied to remote sensing applications, which overcome the drawback of traditional techniques. In this research, a low-cost UAV-based multi-sensor data fusion model was developed for land cover classification based on a DCNN. For the purpose of this research, two datasets were collected at two different urban locations using two different UASs. A DCNN model, based on U-net with Resnet101 as a backbone, was used to train and test the fused image/LiDAR data. The maximum likelihood and support vector machine techniques were used as a reference for classifier comparison. It was shown that the proposed DCNN approach improved the overall accuracy of land cover classification for the first dataset by 15% compared to the reference classifiers. In addition, the overall accuracy of land cover classification improved by 7%, and the precision, recall, and F-measure improved by 18% when the fused image/LiDAR data were used compared to the images only. The trained DCNN model was also tested on the second dataset, and the obtained results were largely similar to those of the first dataset.

Integration of Global Positioning System (GPS) and Inertial Navigation System (INS) integrated system involves nonlinear motion state and measurement models. However, the extended Kalman filter (EKF) is commonly used as the estimation filter, which might lead to solution divergence. This is usually encountered during GPS outages, when low-cost micro-electro-mechanical sensors (MEMS) inertial sensors are used. To enhance the navigation system performance, alternatives to the standard EKF should be considered. Particle filtering (PF) is commonly considered as a nonlinear estimation technique to accommodate severe MEMS inertial sensor biases and noise behavior. However, the computation burden of PF limits its use. In this study, an improved version of PF, the unscented particle filter (UPF), is utilized, which combines the unscented Kalman filter (UKF) and PF for the integration of GPS precise point positioning and MEMS-based inertial systems. The proposed filter is examined and compared with traditional estimation filters, namely EKF, UKF and PF. Tightly coupled mechanization is adopted, which is developed in the raw GPS and INS measurement domain. Un-differenced ionosphere-free linear combinations of pseudorange and carrier-phase measurements are used for PPP. The performance of the UPF is analyzed using a real test scenario in downtown Kingston, Ontario. It is shown that the use of UPF reduces the number of samples needed to produce an accurate solution, in comparison with the traditional PF, which in turn reduces the processing time. In addition, UPF enhances the positioning accuracy by up to 15% during GPS outages, in comparison with EKF. However, all filters produce comparable results when the GPS measurement updates are available.

Pavement surface maintenance is pivotal for road safety. There exist a number of manual, time-consuming methods to examine pavement conditions and spot distresses. More recently, alternative pavement monitoring methods have been developed, which take advantage of unmanned aerial systems (UASs). However, existing UAS-based approaches make use of either image or LiDAR data, which do not allow for exploring the complementary characteristics of the two systems. This study explores the feasibility of fusing UAS-based imaging and low-cost LiDAR data to enhance pavement crack segmentation using a deep convolutional neural network (DCNN) model. Three datasets are collected using two different UASs at varying flight heights, and two types of pavement distress are investigated, namely cracks and sealed cracks. Four different imaging/LiDAR fusing combinations are created, namely RGB, RGB + intensity, RGB + elevation, and RGB + intensity + elevation. A modified U-net with residual blocks inspired by ResNet was adopted for enhanced pavement crack segmentation. Comparative analyses were conducted against state-of-the-art networks, namely U-net and FPHBN networks, demonstrating the superiority of the developed DCNN in terms of accuracy and generalizability. Using the RGB case of the first dataset, the obtained precision, recall, and F-measure are 77.48%, 87.66%, and 82.26%, respectively. The fusion of the geometric information from the elevation layer with RGB images led to a 2% increase in recall. Fusing the intensity layer with the RGB images yielded a reduction of approximately 2%, 8%, and 5% in the precision, recall, and F-measure. This is attributed to the low spatial resolution and high point cloud noise of the used LiDAR sensor. The second dataset crack samples obtained largely similar results to those of the first dataset. In the third dataset, capturing higher-resolution LiDAR data at a lower altitude led to improved recall, indicating finer crack detail detection. This fusion, however, led to a decrease in precision due to point cloud noise, which caused misclassifications. In contrast, for the sealed crack, the addition of LiDAR data improved the sealed crack segmentation by about 4% and 7% in the second and third datasets, respectively, compared to the RGB cases.