Ali Hosseininaveh Ahmadabadian

Motorcycle detection and collision warning are essential features in advanced driver assistance systems (ADAS) to ensure road safety, especially in emergency situations. However, detecting motorcycles from videos captured from a car is challenging due to the varying shapes and appearances of motorcycles. In this paper, we propose an integrated and innovative remote sensing and artificial intelligence (AI) methodology for motorcycle detection and distance estimation based on visual data from a single camera installed in the back of a vehicle. Firstly, MD-TinyYOLOv4 is used for detecting motorcycles, refining the neural network through SPP (spatial pyramid pooling) feature extraction, Mish activation function, data augmentation techniques, and optimized anchor boxes for training. The proposed algorithm outperforms eight existing YOLO versions, achieving a precision of 81% at a speed of 240 fps. Secondly, a refined disparity map of each motorcycle’s bounding box is estimated by training a Monodepth2 with a bilateral filter for distance estimation. The proposed fusion model (motorcycle’s detection and distance from vehicle) is evaluated with depth stereo camera measurements, and the results show that 89% of warning scenes are correctly detected, with an alarm notification time of 0.022 s for each image. Outcomes indicate that the proposed integrated methodology provides an effective solution for ADAS, with promising results for real-world applications, and can be suitable for running on mobility services or embedded computing boards instead of the super expensive and powerful systems used in some high-tech unmanned vehicles.

In this article, a method for the extrinsic calibration of a 2D laser range finder and a camera is presented. This technique produces a 3D point cloud from the test field by connecting a laser range finder to a servomotor. In this study, ping balls and standard photogrammetric targets were employed as a test field. Ping pong balls are used because they can be easily recognized in data from laser range finder and camera. To calculate extrinsic calibration parameters between a camera and laser range finder, these balls are employed as control points in the data. The extrinsic calibration of the Laser range finder and camera is carried out using the point cloud created from the test field and the photos captured from the test field. In this method, a sphere is fitted to each ping pong ball's points in the 3D point cloud, and the coordinates of that sphere's center are taken to be the coordinates of that ball. By measuring the distances between various targets in the test field, the scale can be resolved. This approach was compared with another state-of-the-art method. The proposed method is more accurate and stable than the alternative way, taking into account the average inaccuracy of check points.

After considering state‐of‐the‐art algorithms, this paper presents a novel method for generating true orthophotos from unmanned aerial vehicle (UAV) images of urban areas. The procedure consists of four steps: 2D edge detection in building regions, 3D edge graph generation, digital surface model (DSM) modification and, finally, true orthophoto and orthomosaic generation. The main contribution of this paper is concerned with the first two steps, in which deep‐learning approaches are used to identify the structural edges of the buildings and the estimated 3D edge points are added to the point cloud for DSM modification. Running the proposed method as well as four state‐of‐the‐art methods on two different datasets demonstrates that the proposed method outperforms the existing orthophoto improvement methods by up to 50% in the first dataset and by 70% in the second dataset by reducing true orthophoto distortion in the structured edges of the buildings.

Imaging network design is a crucial step in most image-based 3D reconstruction applications based on Structure from Motion (SfM) and multi-view stereo (MVS) methods. This paper proposes a novel photogrammetric algorithm for imaging network design for building 3D reconstruction purposes. The proposed methodology consists of two main steps: (i) the generation of candidate viewpoints and (ii) the clustering and selection of vantage viewpoints. The first step includes the identification of initial candidate viewpoints, selecting the candidate viewpoints in the optimum range, and defining viewpoint direction stages. In the second step, four challenging approaches—named façade pointing, centre pointing, hybrid, and both centre & façade pointing—are proposed. The entire methodology is implemented and evaluated in both simulation and real-world experiments. In the simulation experiment, a building and its environment are computer-generated in the ROS (Robot Operating System) Gazebo environment and a map is created by using a simulated robot and Gmapping algorithm based on a Simultaneously Localization and Mapping (SLAM) algorithm using a simulated Unmanned Ground Vehicle (UGV). In the real-world experiment, the proposed methodology is evaluated for all four approaches for a real building with two common approaches, called continuous image capturing and continuous image capturing & clustering and selection approaches. The results of both evaluations reveal that the fusion of centre & façade pointing approach is more efficient than all other approaches in terms of both accuracy and completeness criteria.

Nowadays, robotic systems such as ground vehicle robots are mostly used in many industrial and military applications. Therefore, the path planning problem in the robotics domain is very important. Moving Obstacles Planner (MOP) algorithms have got the researchers interests in recent years and some of the most recent ones have been implemented in Robot Operating System (ROS) which is an open source middle wear to work with robots. This paper aims to compare the state-of-the-art MOP algorithms including Rapidly exploring Random Tree (RRT) and those implemented in the ROS navigation stack such as Dynamic Window Approach (DWA) local planner coupled with Dijkstra and A* as global planners on a six-wheeled robot known as MOOR in simulation environment. The results reveal that all of these algorithms have been designed for a square shape footprint robot and thus have limitations for MOOR with a rectangular footprint shape.

Purpose: This paper aims to propose an automatic imaging network design to improve the efficiency and accuracy of automated construction progress monitoring. The proposed method will address two shortcomings of the previous studies, including the large number of captured images required and the incompleteness and inaccuracy of generated as-built models. Design/methodology/approach: Using the proposed method, the number of required images is minimized in two stages. In the first stage, the manual photogrammetric network design is used to decrease the number of camera stations considering proper constraints. Then the image acquisition is done and the captured images are used to generate 3D points cloud model. In the second stage, a new software for automatic imaging network design is developed and used to cluster and select the optimal images automatically, using the existing dense points cloud model generated before, and the final optimum camera stations are determined. Therefore, the automated progress monitoring can be done by imaging at the selected camera stations to produce periodic progress reports. Findings: The achieved results show that using the proposed manual and automatic imaging network design methods, the number of required images is decreased by 65 and 75 per cent, respectively. Moreover, the accuracy and completeness of points cloud reconstruction is improved and the quantity of performed work is determined with the accuracy, which is close to 100 per cent. Practical implications: It is believed that the proposed method may present a novel and robust tool for automated progress monitoring using unmanned aerial vehicles and based on photogrammetry and computer vision techniques. Using the proposed method, the number of required images is minimized, and the accuracy and completeness of points cloud reconstruction is improved. Originality/value: To generate the points cloud reconstruction based on close-range photogrammetry principles, more than hundreds of images must be captured and processed, which is time-consuming and labor-intensive. There has been no previous study to reduce the large number of required captured images. Moreover, lack of images in some areas leads to an incomplete or inaccurate model. This research resolves the mentioned shortcomings.