On-Board, Real-Time Preprocessing System for Optical Remote-Sensing Imagery
With the development of remote-sensing technology, optical remote-sensing imagery processing has played an important role in many application fields, such as geological exploration and natural disaster prevention. However, relative radiation correction and geometric correction are key steps in preprocessing because raw image data without preprocessing will cause poor performance during application. Traditionally, remote-sensing data are downlinked to the ground station, preprocessed, and
... uted to users. This process generates long delays, which is a major bottleneck in real-time applications for remote-sensing data. Therefore, on-board, real-time image preprocessing is greatly desired. In this paper, a real-time processing architecture for on-board imagery preprocessing is proposed. First, a hierarchical optimization and mapping method is proposed to realize the preprocessing algorithm in a hardware structure, which can effectively reduce the computation burden of on-board processing. Second, a co-processing system using a field-programmable gate array (FPGA) and a digital signal processor (DSP; altogether, FPGA-DSP) based on optimization is designed to realize real-time preprocessing. The experimental results demonstrate the potential application of our system to an on-board processor, for which resources and power consumption are limited. on-board processing can reduce the cost and complexity of ground processing systems and solve the delay problem in image acquisition, analysis, and application. The acquired remote-sensing images may contain uneven radiation brightness stripes and deformation areas, due to the defects of the sensors and the relative movement between satellite platforms and the Earth    . Therefore, the acquired raw data from sensors on satellite platforms cannot be used directly. So, image preprocessing is a necessary step to solve such crucial problems. There are several necessary steps for preprocessing within charge coupled device (CCD) camera images, such as relative radiation correction (RRC), geometric correction (GC), and multi-CCD stitching (MCCDS). Numerous studies have been performed to satisfy the needs of on-board processing. Cong Li et al.  introduced a new volume calculation formula and developed a new real-time implementation of a maximum simplex volume algorithm, which is suitable for real-time, on-board processing. Qian Du et al.  employed a small portion of pixels in the evaluation of data statistics to accelerate the real-time implementation of detection and classification. This design achieved fast, real-time, on-board processing by reducing computational complexity and simplifying hardware implementation. Scholars have also conducted related studies of architecture implementation and efficient algorithm mapping. El-Araby et al.  presented a reconfigurable computing real-time cloud detection system for satellite on-board processing. Kalomiros et al.  designed a hardware/software field-programmable gate array (FPGA) system for fast image processing, which can be utilized for an extensive range of custom applications. Winfried et al.  designed an on-board, bispectral infrared detection system, which is based on the neural network processor NI1000, a digital signal processor (DSP), and a FPGA. The system can perform on-board radiometric correction, geometric correction, and texture extraction. Botella et al.  proposed an architecture for a neuromorphic, robust optical flow based on a FPGA, which was applied in a complicated environment. Multi-core processors and graphic processing units (GPUs) for achieving real-time performance of the Harsanyi-Farrand-Chang (HFC) method for a virtual dimensionality (VD) algorithm was proposed for unmixing  . Carlos et al. presented the first FPGA design for the HFC-VD algorithm to realize unmixing  . The previously mentioned methods-GPU, FPGA, and DSP-are the most common processors for implementing these algorithms in real time. In a ground processing system, a GPU is the popular choice for a preprocessing system. Although a GPU can provide high computing performance, it consumes considerable energy and cannot achieve the radiation tolerance required for an on-board environment. Therefore, a GPU cannot be adapted to an on-board processing system. To satisfy the requirements of on-board processing, this system should be implemented using a FPGA, which has low power consumption and high radiation resistance    . Considering the computational complexity of a preprocessing algorithm, the use of a DSP as a co-processor is common to perform processes that are not computationally demanding and need to be sporadically executed. Although some publications have designed GC systems based on a FPGA, these systems are not suitable for remote-sensing images    or cannot achieve the complete process  . To the best of our knowledge, no such hardware systems have been proposed for remote image preprocessing, probably because of the complex computations and data management required. However, such a preprocessing step should be executed on this platform to achieve higher performance. The process of image preprocessing can be decomposed into two parts. The first step calculates the model parameters. This step processes small amounts of data but involves complex calculations (such as sine and cosine functions), making it suitable for a DSP. The second step uses the model parameters to perform a pixel-by-pixel, gray-scale calculation and obtain the output image. When the pixels are calculated in this step, parallel calculations are appropriate, because the calculation forms of all the pixels are similar. However, due to the irregularity of the image deformation and other issues, there are several problems in the pixel calculation step. First, the calculation of each pixel coordinate requires many parameters and a large amount of hardware computing resources.