System Design And Image Mosaicing For Array Aerial Reconnaissance CCD Camera

As one of the most important equipments in aerial reconnaissance mission, aerial reconnaissance cameras have been paid a great deal of attention by each air force in the word. From the 21st century, modern combat theories and new weapons are spring up with many new techniques, and these propose more technical requirem-ents for aerial reconnaissance cameras. So, the aerial reconnaissance capability of our country will be greatly improved by researching on a new array aerial reconaissance CCD camera (AARCC), which can meet the requirements of aerial reonnaissance mission at this stage. This dissertation comes from the research subject -"One Array Aerial Reconnaissance CCD Camera System", and it maily forcous on the requirements of image quality and the field of view (FOV) in this camera. Image motion, image overlap rate fluctuation and image mosaicing are studied on the basis of canmera system design, and it is desgined by single CCD sensor scanning mechanism. The main contents of this dissertation are as follows.An AARCC of horizontal scanning structure is first designed in this dissertation, which bases on existing AARCC system and former film aerial reconnaissance design experience. Ideal camera model is then derived by the work principle of camera, transform theory of central projection, and suitable coordinates. Expression of camera scanning parameters and image overlap rates is also derived in follow. Calculation results of the FOV have proved the efficiency of this camera system design, and it can meet the requirement of aerial reconnaissance mission in super-wide-angle.Forward and attitude disturbance image motion models are separately derived in following chapters, and they are based on the factors of image motion and the ideal camera image model. Compensating systems of these image motions are designed by the quantitative analysis results between image motion of disturbance and photogra-phy angle. One of them is designed by electronic manner to remove forward image motion, and the other is designed by optical manner to remove attitude disturbance image motion in roll and pitch direction. A group of images taken during the flying test demonstrate that image motion is effectively restrained by those compensating system, and the image quality can satisfy the system design requirements.2D transform model of image sequences is then derived in homogeneous coordinates, which conforms on the disturbance affection to the image overlap rates of this AARCC. Relationship between disturbance and image overlap rates in two directions are researched by this model. Parameters of scanning gaps, time intervals and scanning numbers is given by the computation results, and fluctuation range of image overlap rates computed by these parameters is also given in this chapter. Image sequences of three channels taken during the flying test indicate that image overlap rates of these images is always existing, which can satisfy the camera design requirements.As the feature of these image sqequences taken by this kind camera is similar to translation transformation, two image mosaicing method are researched by an image smooth algorithm of bilinear interpolation strategy, which is designed according to the procedure of image mosaicing. One of them is registrated by object edge features, and the other is based on templates matching. In the former method, Canny opterator and Hough transform are fisrt used to extract linear edge features of the obejects in these images, and intersections of these features are used to complete the process of image mosacing. Experimental results verify that this method can supply high image registration precision and excellent image mosaicing ability when suitable targets are existing in the overlap areas. To make up the disadvantage of this method in application of aerial reconnaissance images, an image mosaicing method of templates matching based on affine transform is proposed, which choose normalize correlation coefficient as the measure of template matching. Despite of its generality, simplicity and low repeatability error, data processing amount of this method is too large and its image registration precision is relating to the matching reference templates. Thus a new image mosaicing method of templates matching based on affine transform and Sobel operator is proposed as well. By this new method, precision of image registration is able to be kept and processing speed of image mosaicing can be improved. Experiment results demonstrate the good popularity and high processing velocity of this method.