Research On Key Technologies For Real-time Acquisition,Transmission And Storage Of Massive Time-series Images From CoaXPress High-speed Cameras

High-speed videogrammetry technology has been widely used in fields such as production inspection,weapon testing,aerospace,and aviation.It can provide non-contact,high-precision,and high spatiotemporal resolution measurement data.With the development of camera sensor technology,high-speed cameras are moving towards larger arrays and higher frame rates.However,this also brings about the challenge of real-time processing of massive data.Real-time and accurate processing of massive image data is crucial for obtaining accurate and reliable data in various engineering experiments.Buffer settings and storage write strategies play a critical role in ensuring real-time collection and storage of massive image data.However,selecting the appropriate buffer settings and storage write strategies remains a challenge.On one hand,the size of the memory buffer is difficult to determine due to the wide variation in the size of high-speed image data,which is dependent on camera parameters.On the other hand,a single storage strategy cannot meet the real-time writing requirements due to the variation in camera frame rates.To address these two issues,this article primarily focuses on the following research work:(1)To address the problem of image loss resulting from inadequate processing in the memory buffer,we designed an adaptive buffer expansion scheme aimed at improving the efficiency of image writing.Firstly,we adopted a FIFO cache algorithm to manage the buffer data and ensure that sequential images were swapped in the buffer in the correct order.Secondly,we designed a circular buffer structure to facilitate the efficient transmission and processing of image sequence data,thereby avoiding data loss or overflow.Thirdly,we dynamically set the buffer size based on the image window and expanded it to the smallest storage unit multiple of the storage medium to fully accommodate the current captured image data and improve writing efficiency.In a comparative experiment between the non-expanded buffer method and the method proposed in this article,the results showed that the proposed adaptive buffer expansion method achieved an average transmission speed of 1.95GB/s and improved the average transmission speed compared to the non-expanded buffer method.(2)To address the issue that a single storage write strategy cannot meet the diverse data writing needs of different frame rates,we propose a storage strategy optimization method based on the storage device IOPS.Firstly,we construct an asynchronous write algorithm combined with the asynchronous I/O model to resolve the problem of thread blocking during high-speed image data writing.Then,we utilize the parameter "r" to determine whether to use the hard disk cache to adapt to various frame rates and storage medium characteristics,thus improving the performance and service life of the hard disk.Through comparative experiments between the single storage strategy(no-cache method and cache method)and the storage strategy proposed in this paper,the results demonstrate that this method can satisfy the bandwidth requirements of high-speed camera writing at different frame rates.When the frame rate is below 8000 fps,the writing speed can exceed 1.9GB/s and is consistent with the no-cache method;when the frame rate is above 8000 fps,the writing speed can reach 0.9GB/s-1.5GB/s and is consistent with the cache method.(3)The reliability of the high-speed video measurement acquisition and storage system proposed in this paper was verified through engineering experiments.Firstly,based on the theory of high-speed video measurement and the characteristics of the spherical shell structure,we completed the configuration of the high-speed video measurement acquisition and storage system and the network sensors.Then,we calculated the dynamic 3D coordinates of the artificial target points in the sequence images through the control point coordinates and target point positioning and tracking algorithm.Finally,during the experiment,the high-speed sequence images acquired by the real-time acquisition and storage system developed in this paper exhibited no frame loss,and the root mean square errors of the displacement results processed in the X,Y,and Z directions were 0.32 mm,0.7 mm,and 0.06 mm,respectively.The first-order frequency of 2.331 Hz obtained by the fast Fourier transform and the first-order frequency of 2.343 Hz calculated by the accelerometer were consistent,validating the high stability of the developed high-speed video measurement real-time acquisition and storage system.