Based on our original development of a new laser absorption spectroscopy chamber (LASC) system, we further report ammonia emission concentration measurement using the LASC system based on deep belief networks (DBNs), aiming to present an effective approach for the retrieval of the gas concentration to increase the measurement accuracy of the LASC system and expand its application to monitoring ammonia emission in farmland. Surrounding the LASC system, an experimental system was constructed, and a DBN algorithm was introduced for gas concentration retrieval. The absorption spectroscopy obtained by the experimental system was first pretreated by an empirical wavelet transform algorithm and principal component analysis method, which greatly improved the signal-to-noise ratio of the signal and reduced the dimensionality of the processed signal to meet the need of the training of DBN model. The results showed that the measured gas concentrations were close to true values with small errors, and the mean relative error obtained by the DBN algorithm (0.37%) was much smaller than those obtained by the back-propagation neural network algorithm (0.97%) and absorbance peak method (2.37%) in a wide range of NH3 standard concentrations. Field experiments verified the effectiveness and reliability of the LASC system when it was applied to ammonia emission measurement in farmland with the concentration retrieval based on the DBN algorithm, which is of importance for its applications in air pollution detection.
Vertical Radial Plume Mapping (VRPM) technique is often used in the measurement of gas emission flux in open space. It is necessary to use optical remote sensing equipment (ORS) to scan multiple measurement points to reconstruct the gas concentration field, but the fluctuation of field environmental conditions and the mechanical error of the system will lead to the optical path deviation. Although the optical path calibration can be completed by researching and positioning the central position of the measurement point according to the signal strength, the search range needs to be preset, which can not balance the time cost and positioning accuracy, reducing the time resolution of the concentration data, and resulting in flux calculation error. To solve this problem, this paper proposes a Q-learning multi-optical path localization method based on detection signal quality. This method uses the change of signal strength when the optical path moves as a reward to learn the environment, affects the selection of the next calibration direction, and makes the optical path preferentially choose the direction with enhanced signal strength. The effectiveness of this method is verified on the 25 * 25 map established of simulating the optical path offset. The results show that this method can get the optimal path to the center point, the minimum number of steps is 14, the running time is less than 2 seconds, and the success rate can reach 100% after many episodes of learning, which proves the effectiveness of Q-learning method in multi-optical path scanning.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.