An Android Device-based Remote Sensing System Of Unmanned Aerial Vehicle(UAV) For Image Acquisition And Analysis

Currently existed remote sensing image acquisition platforms for UAV were comprised of varieties of devices, which were hard to develop, large in weight, high in expenses and with high requirement of their mounting UAV leading to difficulties in transforming researches into actual practices. Hence, an Android smart device based remote sensing image acquisition platform for UAV was developed in this research. Since the light weight and small size of the Android smart devices, it lowers the volume and cost of their mounting UAVs and lengthens their flight endurances. What's more, since Android smart devices are of high popularity, low cost, easy in maintenance and migration for relevant applications, which is convenient for the promotion of the research achievements. The contents and results of this research are described as follow:(1) An Android application developed with Android SDK, Eclipse Integrated developed environment(IDE) and the JAVA language, a ground PC software developed with Labwindows/CVI IDE and the C language, an external device extension component developed with 8051 micro-processing unit(MCU) and HY-05 Bluetooth adapter were deveopled respectively. Wireless network was used to transmit not only the preview frames, data obtained form the built-in seneors and GPS of the Android smart devices to the ground PC terminal, but also the controlling command from the ground PC terminal to the Android terminal. In actual practices, such designated functions together with the controlling and internal state access of an ADC Lite multispectral camera connected to the external device extension component were achieved;(2) The compression algorithms of the RGB image were compared. Compression algorithms were applied to the preview frames transmitted from the Android terminal to the ground PC terminal to increase the velocity of the preview frame transmission and reduce the possibilities of frame loss, frame suspension and screen blur which enhance the reliability and the real time performance. Experiments of differences of transmission velocity between the YUV420 sp coded and JPEG coded preview frames in different resolution under both static and dynamic scene fields were conducted. The results indicated that the JPEG coded preview frames were with higher transmission velocity than the YUV420 sp coded frames and showed extremely significant differences in the experiment results calculated with one-way ANOVA test. What's more, applying JPEG coding format can reduce the possibilities of frame loss and suspension;(3) The calibration algorithms for the distorted images were studied. Applying calibration algorithms to the acquired images can restore the perspective distortion due the flight posture of the UAVs together with the shearing distortion and lens distortion of the photogrammetric cameras. Calibration algorithms used for the restoration of either barrel distortion or perspective distortion were raised respectively and applied to the experiment images. The results indicated that the calibration algorithms for barrel distortion lowered the vertical distortion carried by both shearing distortion and lens distortion and showed extremely significant differences in the experiment results calculated with one-way ANOVA test. A calibration template image was used to conduct the experiment of perspective distortion calibration and showed good results with the parameters of the Hough transformation;(4) Different algorithms of image registration were compared. Geographic information extracted from the Geo TIFF format coded remote sensing images were used for the image registration together with the coordinate transformation algorithms; Fourier transformation-based phase correlation algorithm was applied to images with translational model for image registration; The SURF algorithm was used for image registration which the separated images are of neither geographic information nor movement models. In accordance with the sensor data obtained from the Android smart devices in actual practices, images acquired from the Android smart devices are not suitable for either Geographic information or phase correlation algorithm based image registration. However, separated images were well registered using SURF algorithm which showed good robustness facing noise, disturbance and scale deformation;(5) Images of cotton field were acquired with the developed image acquisition system and related mathematical models were established with extracted and screened feature values obtained from the segmented cotton boll regions. Images of unknown cotton field were used for model verification which the recognition rates between the RGB values and linear regression models were: B-G: 65.98%; B-R: 64.89%; R-G: 75.78%; the recognition rates between the texture parameter values of the gray level co-occurrence matrix and quadratic regression models were: pixel distance-ASM: 98.11%; pixel distance-contrast: 98.89%; pixel distance-correlation: 97.78%; pixel distance-mean: 6.99%; pixel distance-sum of average: 14.76%.Above all, the image acquisition and analysis platform developed in this study was able for image acquisition, image calibration and registration to the acquired image and the retrieve of agriculture information from those images, which showed feasibility in actual practices.