Zygmunt Kuś

Object Tracking in a Picture during Rapid Camera Movements

Unmanned aerial vehicles (UAV) are very useful platforms for detecting and tracking objects which are located on the ground. The crux of the considered problem is the tracked object which disappears from a field of view. The above mentioned may be caused by rapid camera movements. One of the examples of such a situation is the camera being attached to helicopter. In case of sudden gust of wind a helicopter trajectory can be considerably and rapidly changed. This results in losing tracked object from the picture we get from the camera. The fundamental idea of the solution, which was presented here in, was based on additional data concerning camera orientation and location. Moreover the distance of a tracked object from the camera is also utilized to correct camera movements.

Helicopter Control Algorithms from the Set Orientation to the Set Geographical Location

The aim of the paper was to develop helicopter control algorithms. On the basis of the results it is possible to control the unmanned helicopter by a person without pilot experience. The suggested solution is based on multilevel control. It allowed to have an effect on helicopter relocation due to setting a requested trajectory. The knowledge mentioned in the literature which concerns the behavior of the helicopter is applied to the synthesis of control system .What is more, the identification of the certain helicopter model elements was made herein. The paper presents the example of the helicopter flight to a set target location. Furthermore, the example provided in the study confirms the correct control system behavior. The possibility of introducing the set location in the coordinate system, which is connected with the ground, allows to reach the set geographical location by the helicopter in an autopilot mode.

Object Tracking for Rapid Camera Movements in 3D Space

The solution of the camera head control problem was presented in the paper. The main goal of developed control algorithm is the object tracking in rapid disturbance conditions. The changes of the helicopter position and orientation during the time of disturbance result in losing tracked object from the field of view. Due to this fact the helicopter control system uses not only the visual information. The essence of the proposed solution is to compute the object position only in such time intervals when the object is in the center of the image. It allows the camera head regulators to compensate change of the camera position and set the camera towards the object. The proposed solution comprises of turning the camera head towards the trucked object in horizontal and vertical planes. The appropriate angles were computed on the basis of rangefinder data (distance between the camera and the object) and GPS and IMU data (the camera position and orientation). Furthermore, examples of the situation when the distortion changes the helicopter position in horizontal and vertical plane were presented in the paper.

Camera Head Control System with a Changeable Gain in a Proportional Regulator for Object Tracking

The goal of the paper was to examine the changeable gain in the camera head control system which is applied to object tracking. The study assumed that in order to track objects moving on the ground, the UAV with the mounted camera would be used herein. During the flight of the UAV, there may appear various disturbances such as rapid rotations and shifts of the UAV. Furthermore, we assumed that the object was moving on the flat ground. The UAV’s velocity was far greater than the maximum velocity of the tracked object. Due to equipping the UAV with flight trajectory control system, it allowed to model the rotations and shifts of the UAV in the particular axes as independent control channels. The behaviour of the control system was also examined for the different values and types of disturbances. Developing the function which defined the relationship between the error signal and gain in the control system let us track the object appropriately. Conclusively, the results indicated on the fact that using changable gains in the camera head control system, allowed to track the object for both, the small and major disturbances. Owing to the changeable gain, it was possible to eliminate rapid reactions of the control system during major disturbances. As a consequence, there was not any overshoot - the object was being tracked even during high values of the disturbances. Simultaneously, it was noticed that the disturbances which shift the UAV in the (x, y) plane, had less impact on the appriopiate object tracking than the disturbances which rotated the UAV in the horizontal or vertical planes.

The Limitation for the Angular Velocity of the Camera Head during Object Tracking with the Use of the UAV

The following paper examined the significance of the velocity limitations in the camera head control system. A prerequisite for the control system was that it should quarntee the tracking of an object moving on the ground, even in the case of disturbances such as the rapid rotations and shifts of the UAV. Equipping the UAV with the flight trajectory control system allowed to model the rotations and shifts of the UAV in the particular axes as independent control channels. What is more, the study examined behaviour of the control system for different values of the disturbances. The paper presented the comparison of the object tracking efficiency for the case with and without the limitations of the angular velocity of the camera head. The results showed that the limitation of the velocity of the camera head rotation are crucial in correct object tracking. There was the relationship between both, the value as well as the type of the disturbance and the requirements regarding the necessary velocity of the camera rotation. The paper showed that at the stage of designing the camera head, it was essential to adjust the velocities of the engines. The rationale for selecting the appropriate engines were the value and type of the disturbances to which the UAV might be exposed during operation. It was suggested in the paper that in the case of the disturbances, which require the velocities of rotation greater than the limit velocities of the engines, it was necessary to use additional information about the tracked object. For example, information about the velocity and direction of the tracked object might be used to track the object despite losing the object from the camera field of view for some time.

The Modified BLT Method for Multivariable Control Systems

The goal of the following paper was the synthesis of multivariable control systems. Due to the popularity and usability of the BLT method for tuning such systems, we proposed a certain modification. The BLT method uses the Ziegler-Nichols (ZN) method for tuning the PI regulators for the main channels. The limitations caused by the ZN method for the plant’s transfer function form resulted in limiting the use of the BLT method. Therefore this paper put forwards the modification based on using the tuning methods different than ZN. This paper shows the examples of using different tuning methods than ZN which were adjusted to the form of the plant’s transfer function. Such an approach allowed to extend the range of the acceptable kinds of the plant’s transfer functions in the main channels during the synthesis of the multivariable control system. It allowed to use the knowledge of single-variable control systems for tuning the multivariable control systems. The example presented in this paper illustrates the usefulness of such an approach.

Suppressing Disturbances in the UAV’s Control System Based on the Modified BLT Method

The paper presents the use of the modified BLT method for the synthesis of the control system which task was to suppress disturbances. The controlled plant constituted a linearized multivariable model for the helicopter-type UAV. The first step of the modified BLT method is to tune PI controllers for each main channel. The article proposed the method of tuning which corresponded to the character and properties of the disturbances. The authors focused on the lowest control level. This method was applied to control the linear velocity of ascending and angular velocities of the UAV’s rotation around the symmetry axes. This study elaborates on the examples of the control system operation for appearing step and sinusoid disturbances. The presented examples illustrated the correctness of the operation of the modified BLT method in the disturbance suppression task during the UAV flight. What is more, the unexpected result was that the modification of the BLT method offered the wider possibilities of PI controllers tuning. This tuning process may take into account various requirements for the control system.

Applying Colour Image-Based Indicator for Object Tracking

The goal of the following paper was to devise a method of object tracking with the application of the tracked objects’ coloured images. The proposed solution was based on the calculation of an indicator which described the colour features of the object. The ratio between the red and green components as well as the ratio between red and blue components were the indicators which defined these features. Moreover, the proposed approach was particularly useful in the cases when the object and terrain colours were significantly different. The abovementioned ratios were used to create the pattern vector. Such a defined pattern vector was used to calculate the error function and the minimum of this function indicated the object location. This paper presented the examples of object tracking for both: the different object colour from the terrain colour and the similar object colour to the terrain colour.

Weighted Pattern Vector for Object Tracking with the Use of Thermal Images

The goal of the following paper was to devise a method of object tracking in a case when we have preliminary knowledge of an object. Tracking the object which state or some properties are known may be conducted with the pattern vector modification basing on this knowledge. The object which has the same visual images and different thermal images, depending on the time the object was moving, may be an example of the abovementioned case. One of such situations, presented as an example in the paper, was the change of the car engine temperature which modified the thermal image whereas the visual image was the same for different engine temperatures. In each moment during object tracking image processing system used thermal and visible images. The authors assumed that we acquire the visual and thermal images in each moment of object tracking. The pattern vectors of the tracked object based on the visual and thermal images were computed separately. The pattern vector and current feature vector for an image of a given type were used to compute the distance between the object pattern vector and feature vector calculated for a given location of the aperture. It was calculated for both: the visual and thermal images. The crux of the proposed method is the algorithm of setting the correct value of the weight with which the visual and thermal pattern vectors are used to calculate the distance. The authors proposed the method based on the knowledge about the object location in its initial position. The pattern vector calculated in this position was treated as a correct and unchangeable one. Assuming that the thermal pattern vector was more useful for object recognition, we proposed the method of increasing weight for a part of the pattern vector which part was based on a thermal image. Finally, this study presented the examples of the object recognition by means of the developed method.

The Method of Developing the Invariant Functions Vector for Objects Recognition from a Given Objects Set

The goal of the following paper was to develop the methodology of recognising a given object out of other objects in the grey scale images. We have presented the method which utilises moment invariants for creating model vectors which define the features of the recognised object. Moreover, we have discussed the rules of creating the model vectors which guarantee differentiation of the given object classes. We have put forward the method of recognising an object in the image. This method is based on searching the points in the image for which there is minimal distance between the model vector and current vector—calculated for each point of the image. On top of that, this study presents the examples of the object recognising by means of the developed method.