| This topic mainly proposes and implements a parking space video monitoring based on embedded devices and parking space detection scheme. Using the embedded network technology and video image processing technology, it provides a more convenient solution to the current parking problem. In addition, this topic also implements the remote system upgrade and maintenance functions based on secure shell protocol (SSH), as well as the system of multiple backup and restore function.In the aspect of software system, this subject has realized the parking video monitoring based on B/S network model and parking space detection based on open source computer vision library (OpenCV). The research object of the detection algorithm is the image in parking video, which acquired by the cameras installed in a fixed position of parking space. First, we should get a preprocessed image of parking spot. Image preprocessing includes image correction, filtering out environmental noise and extracting a ROI area of the car. Then,we can select a detection area of parking space, on the basis of the pixel threshold of the various prior offline conditions. By using OpenCV image processing function under the embedded Linux OS, we realize the detection of parking space, and save this results to the embedded database. Then we should send the parking information to the PC or hand tool, and display it in the form of images, which ultimately achieves the purpose of detecting remote parking space.Parking state detection uses three criteria:One is the difference average of parking spaces; The second is the linear correlation of the backgroud parking image and the current parking image; The third is the edge points of the parking area. The results show that the scheme can run well in the embedded platform, and provide support for parking guidance.In the aspect of hardware system, the system adopts processor DM3730 which is used in ARM and DSP. DM3730 is a high performance., enhanced digital media processor which integrates up to the ARM Cortex-A8 processor architecture up to 1 GHz and DSP core up to 800 MHz with advanced digital signal processing. As is known to all, DSP plays an effective part in the field of graphics and images, such as 2D and 3D image processing, image compression and transmission, image enhancement, animation and digital-map, robot-vision, mode-identification and workstations, etc. So this system gives the complex job of the image preprocessing part to DSP processing, such as removing the yawp in the image and transforming RGB to HSI, etc.Although Linux has many superiority as an embedded OS, because of the feature of the embedded operating system itself, such as the variety of the embedded application range, which causes the maintenance problems of embedded product to become more difficult and prominent. After embedded system is used in the actual environment of operation, the unexpected disaster such as power-done, user error or maliciously modify, delete data, and part error that cannot be full tested in the development and will lead to function failure, even can cause system paralysis. If using artificial regeneration, sometimes it is not convenient due to the factors such as installation location. Therefore, automatically backup and recovery mechanism of embedded system is the guarantee of the easiest and most useful way to recover data, and gradually become a problem in the application of embedded systems. In this paper, some problems about backup technology exist in the current embedded system. Through the research of embedded U-Boot and the Linux kernel, we backup several embedded system image on a set of equipment, and only one system runs every time. When a operating system fail to run, it’ll trigger the system recovery mechanism, use the next available backup partition overwrite the failure partition, then start the next available systems. The system does not need backup device, greatly saves cost and power consumption, and improves the dependability of the system. |
PDF Full Text Request