In aerial image object detection, how to efficiently detect different size objects in input images of different scales and obtain a unified multi-scale representation of the object is an important issue. Existing methods rarely consider the connection between multi-scale training and multi-scale inference, and do not well optimize the constraint of input object samples in the multi-scale training process, which limits the performance of multi-scale representation. In this study, an efficient object detection algorithm for aerial images is proposed to alleviate this problem. Firstly, we propose to use metric learning to obtain the scale representation boundary of each object class, reduce the support of indistinguishable objects at extreme scales in the training process, and enhance the effect of multi-scale representation. Secondly, indistinguishable small objects are merged into small object regions, and these regions are trained to recommend the detector to detect small objects on the following high-resolution scale. Thus, a reasonable association between multi-scale training and inference is established, and the efficiency of multi-scale inference is considerably improved. The proposed algorithm has been tested on three popular aerial image datasets, including VisDrone, DOTA and UAVDT. Experimental results show that it can improve the detection accuracy and reduce the number of processing pixels.

This paper presents a nonlinear model of the quadcopter to simulate the actual motion. An aerodynamic model of the rotor is obtained through wind-tunnel tests, which consider the variations in the rotor’s angular velocity, incoming flow speed, and angle of attack of the rotor ( αp ). The quadcopter’s fuselage aerodynamics and mutual interference are simulated using the computational fluid dynamic (CFD) method. The mesh motion method is used to simulate the rotation of the rotor. It can be found that the interference has a significant impact on the aerodynamic load of the fuselage and the thrust and pitch moment of the rear rotor. The mathematical expression of the interference model is given by adopting the idea of the small disturbance hypothesis. The yaw dynamics model is also improved by considering the motor’s dynamics. The quadcopter’s nonlinear model can be obtained by combining the rotor model, fuselage model, motor model, and interference model. Then, flight experiments are done at hovering and forward flight states to verify the established model, and the experiment results are in good agreement with the nonlinear model. Finally, the quadcopter’s characteristic descriptive abilities of different accurate models are discussed to help designers to build a suitable quadcopter model according to the requirements.

Aggressive flight has become increasingly important for expanding the applications of quadrotors. The typical characteristic of large and rapid changes in commands poses stringent demands on the maneuverability of quadrotors. Ensuring flight stability alone is not enough; dynamic responses must also be selectively constrained, presenting quadcopter flight control with daunting challenges. The prescribed performance control (PPC) method is seen as having the potential to solve this problem by allowing for the constrained control of specified performance, leading to extensive research. However, its practical application still faces challenges, such as the system divergence caused by errors exceeding boundaries due to sudden command mutations. This paper presents a robust dynamic event-triggered PPC (DETPPC) method for an aggressive quadrotor flight. By assessing the direction and proximity of tracking errors approaching constraint boundaries, a dynamic event-triggered compensation mechanism for performance function boundaries is established to mitigate the divergence caused by error surpassing and to preserve preset control over the targeted metrics. Controllers were designed for both the translational and rotational subsystems of the quadrotor, and stability analysis was conducted based on Lyapunov functions. Simulation tests on agile trajectory tracking and abrupt attitude control were carried out, demonstrating the effectiveness of the proposed method.