Control and signal processing system of hyperspectral FLS LiDAR Viktor Alekseyev, Sergey Babichenko, Innokenty Sobolev Laser Diagnostic Instruments AS (LDI), Kadaka tee 113A, 12915 Tallinn, Estonia, E-mail:
[email protected] ABSTRACT: The FLS®-series - a family of hyperspectral fluorescent LiDARs - provide real-time data acquisition from complex environments for use in multi-tier environmental and resource surveys. The FLS® stands for Fluorescent Light Detection And Ranging (LiDAR) System. The FLS®-LiDARs are aimed at pollution monitoring of terrestrial, river, lake and marine targets, oil pipeline leak detection and oil exploration. The FLS® series include longsensing distance airborne models, and short-distance shipborne or ground-based models. The instrument is controlled using Windows XP/Vista based software operating on a standard notebook computer. The typical sensitivity of device is 1 ppm substance in water or 1 mg/m2 on ground. The device is also sensible to detect microleakages in oil or gas pipelines.
Mbit of data. The data acquisition system must be able to receive all LIF spectra, make storage of data and process the data in real-time to identity the target. The control system is responsible for all peripheral devices simultaneous work and system error handling. These requirements to hyperspectral LIF imaging system are hurdles of design of such sort of applications.
1 Introduction Laser induced fluorescence (LIF) spectroscopy is a powerful, noncontact tool for the analysis of natural targets in complex environments. LIF analyzers are based upon the capture and analysis of spectra of fluorescence response, induced in the target object by illumination with monochromatic laser emission at one or more wavelengths. LIF reading is a composite of fluorescence responses of individual compounds present in the target object (Fig. 1). Such molecular responses are deterministic for many compounds. As a result, many objects of study have a specific shape of LIF spectra, which are used for remote identification and characterization of sensed object (Fig. 2[1]). The intensity of fluorescence depends on substance concentration and is used for quantitative analysis of the sensed object [2][3]. The principal merit of FLS® LiDAR is its hyperspectral receiving system. Hyperspectral detection refers to simultaneous recording the comprehensive shape of LIF spectrum per each laser pulse. At such detection every laser spot on the target has an associated continuous spectrum which is analyzed. The hyperspectral detector of such LIDAR consists typically of a polychromator, defining detection spectral range and resolution, and a multichannel gated optical detector to read-out LIF spectrum [4]. One of the unique aspects of this form of acquisition system is the massive amount of data that it generates. Because every single read-out of the detector is taken 270 times over one second on 512 pixels (0.58 nm per pixel), the result of one second scanning contains 2.2
Fig. 1. LIF spectra of various surfaces
Fig. 2. Schematic diagram of LIF spectra used for substance identification
2 Control and signal processing system description A common need of control and signal processing instruments is a hard real-time operation system (RTOS).
Fig. 3. Schematic diagram of hyperspectral LIF imaging application
Fig. 4. Simplified microprocessor main routine In LiDAR application the order of time delay between laser pulse and induced fluorescence of target lies in nanosecond range, which gives a sufficient complex challenge to perform the required tasks before the deadline. On the other hand in hyperspectral LIF imaging application, where the hard real-time operation can be
achieved by microprocessor system, the operator environment, flexibility, data process, and large managed storage needs are forcing to use Windows based software operating on notebook or desktop PC. The connectivity and simultaneous work of combination microprocessor -
PC software can be achieved by detailed analysis of whole application. The general solution for such type of application is provided on Fig. 3. The hyperspectral FLS® LiDAR has two types of peripheral devices. The main parts of assembly are eximer or dye laser (sensing sources) and Multichannel Optical Detector (MOD) coupled with spectral unit, the secondary are accessorial devices like GPS, laser range finder, camera and scanner, which role is to upgrade the whole system and obtain additional functionality. The microprocessor here is a self-inclusive synchronising device. It is driven by MOD synchro signal interrupt request, which is also a main clock signal for the internal microprocessor cycle.
Fig. 5. Time diagram of MOD One hyperspectral data array consists of three measured spectra: before the laser pulse, at the laser pulse and after the laser pulse. Spectra before and after the laser pulse are background data of the same point of target. It is typical that these spectra hold different constituent of noise like sun flares, temperature noise and a constant component of signal which can be excluded in further from induced fluorescence spectrum by subtraction. The microprocessor after each synchro signal must be ready to receive 512 data samples from MOD (Fig. 5); each sample size is two bytes, independently from other currently running internal tasks. This request is achieved by the data bus Fast Interrupt Request (FIQ) maximum priority service. The FIQ service is written on optimised assembly language. The hard real-time operating system, which was used in project “LiDAR for Air Estonia” is shown on Fig. 4. No scheduling algorithm was implemented in given system; however the tasks were divided by priorities in hardware vector interrupt controller and to one of them were given the fast interrupt service level, which is specific feature for this type of microprocessor system. This system has two hard real-time constraints – the MOD “data on bus ready” signal and Image intensifier (DEP) strobe signal. As it was already mentioned the order of time delay between laser pulse and backed induced fluorescence signal is in nanosecond range. In example, for the common helicopter altitude of 150 m, this time is 1000 ns, for the altitude of 160 m it is 1067 ns. As a result the precision of delay is determinate by the speed of light, for negative 6 ns delay mistake the actual light will be taken from positive 1 m wrong altitude relative to the ground. Actually this problem is solvable by using the Image intensifier “shutter window”, like
exposition in cameras, and uninterruptible assembly written microprocessor precision routine.
3 Data stream description The data stream is a complex of procedures, data declarations and transport layers. The stream includes the data exchange between embedded system and PC and the Windows based software data interchange. The data between devises is transported by Ethernet physical connection in two types of datagrams: IP/TCP and IP/UDP. The IP/TCP is applied in system data exchange layer, like operator commands and scheduled status verification procedure when there is no ongoing spectral data processing. The IP/UDP datagram is used to transmit the full block of information including status and spectrum. As a result when data processing is on, the full microprocessor data is transmitted on one atomic operation. It can be arguably to use the insecure UDP protocol datagrams. However, in current application, where the main role is given to the fast data acquisition system, unnecessary acknowledgments and data retransmission on transport layer are unacceptable. The receiver must handle the incoming data without disturbing the sender, which is similar to video or audio retransmission. The data packet or “state”, which is described by the structure in common header file, includes hardware and system information with the spectral data. So to the each spectrum in state structure is handled corresponding system information. This gives the capabilities to extract the spectral data in bond with system information and gain a full picture of controller status on the image receiving time moment, for example every LIF spectrum has its own GPS world coordinates and scanner relative position. The state size is selected according to the maximum UDP datagram size [5][6], as it gives the opportunity to free the upper transport layers in microprocessor transfer system from the data fragmentation procedure. To maintain such small data size in outgoing packets, the procedure of data conversion is build in every microprocessor peripheral routine, like the NMEA GPS string is converted to binary format in GPS processing subtask (Fig. 4). The complex Windows based software, written as MFC application, consists of next parts: 1. Hardware compatibility layer; 2. Advanced control and debug layer; 3. Operator control; 4. Real-time data processing and vision system; 5. Storage database procedure; 6. Post-processing procedure; The state between these layers is shifted by two types of transfer: direct data cyclic buffer read pointer transfer and via the events mechanism. The read pointer is available only on low software layers, due to their purpose and functionality. To achieve a fast and uninterrupted data exchange between microprocessor and PC systems the application
has a build in RAM cyclic buffer for 5000 full state structures (Fig. 6).
and until it would not get through all subscribers the thread is halted. To reduce the graphical and computational load on the core, the data processing and visualisation system have dynamic thread priority. The visualization system load is controlled by changing the frame rate of displayed spectra as well.
4 Conclusions and future work Introduced control and signal processing system has evolved for many years. Modern technologies allow receiving or improving system functionality without reducing its complexity. Future development will be directed on more flexible and easy customised solution, to be able to use different types of lasers and many dimensional hyperspectral LIF data. Fig. 6. Windows application data interchange
References [1]
Applying application layer division into the software allows offloading the core of PC. However, when the system was tested on present-day dual core computer, the load indications were high and software performance was unsatisfactory. Leaning on multitasking feature of Windows OS, the application itself has multitasking or multithreading approach. Such approach eliminates the standby time for some critical parts of application.
[2]
[3]
[4]
[5]
[6] Fig. 7. Multithreading diagram of LiDAR PC software The thread diagram of LiDAR PC software is given in Fig. 7. The main thread is the base of application, where all objects and dialog windows are constructed and redraw. The low level treads are: receive data thread and handle data thread. The receive data threads only purpose is to receive non-stop incoming data. The thread priority is set to the highest level, as it allows fastest data processing and lowest response time. The handle data thread checks the RAM buffer status. When there is nontreated data in buffer the handle data thread raises event to subscribed classes. The data is sent to application layers
M. Lennon, S. Babichenko, N. Thomas, V. Mariette, G. Mercier and A. Lisin (2006). Detection and mapping of oil slicks in the sea by combined use of hyperspectral imagery and laser induced fluorescence. EARSeL eProceedings, 5 (1): 120128. S. Babichenko, L. Poryvkina, V. Arikese, S. Kaitala, H. Kuosa (1993) Remote sensing of phytoplankton using laser induced fluorescence. Rem Sens Environ 45:43- 50. S. Babichenko, A. Dudelzak and L. Poryvkina (2004). Laser Remote Sensing of Coastal and Terrestrial Pollution by FLS Lidar, EARSeL eProceedings 3 (1): 1-7. A.E. Dudelzak, S.M. Babichenko, L. Poryvkina and P. Wassmann (2001). Spectroscopic Lidar Sensing in Three-Dimensional Profiling of Marine Organics, International Journal of Remote Sensing, 22 (2&3): 385-402. C. Horing, "A Standard for the Transmission of IP Datagrams over Ethernet Networks", RFC-894, Symbolics Cambridge Research Center, April 1984. J. Postel, "User Datagram Protocol", RFC-768, UCS/Information Sciences Institute, August 1980.
