Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review
Underwater biotope mapping: automatic processing of underwater video data. / Iakushkin, O. O.; Pavlova, E. D. ; Lavrova, A. K. ; Polovkov, V. V. ; Frikh-Khar, A. U. ; Pen, E. A. ; Terekhina, Y. E. ; Bulanova, A. A. ; Shabalin, N. V. ; Sedova, O. S. .
The 6th International Workshop on Deep Learning in Computational Physics (DLCP2022). 2022.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review
}
TY - GEN
T1 - Underwater biotope mapping: automatic processing of underwater video data
AU - Iakushkin, O. O.
AU - Pavlova, E. D.
AU - Lavrova, A. K.
AU - Polovkov, V. V.
AU - Frikh-Khar, A. U.
AU - Pen, E. A.
AU - Terekhina, Y. E.
AU - Bulanova, A. A.
AU - Shabalin, N. V.
AU - Sedova, O. S.
PY - 2022/12/6
Y1 - 2022/12/6
N2 - The task of analysing the inhabitants of the underwater world applies to a wide range of applied problems: construction, fishing, and mining. Currently, this task is applied on an industrial scale by a rigorous review done by human experts in underwater life. In this work, we present a tool that we have created that allows us to significantly reduce the time spent by a person on video analysis. Our technology offsets the painstaking video review task to AI, creating a shortcut that allows experts only to verify the accuracy of the results. To achieve this, we have developed an observation pipeline by dividing the video into frames; assessing their degree of noise and blurriness; performing corrections via resolution increase; analysing the number of animals on each frame; building a report on the content of the video, and displaying the obtained data of the biotope on the map. This dramatically reduces the time spent analysing underwater video data.Also, we considered the task of biotope mass calculation. We correlated the Few-shot learning segmentation model results with point cloud data to achieve that. That provided us with a biotope surface coverage area that allowed us to approximate its volume. Such estimation is helpful for precise area mapping and surveillance.Thus, this paper presents a system that allows detailed underwater biotope mapping using automatic processing of a single camera underwater video data. To achieve this, we combine into a single pipeline a set of deep neural networks that work in tandem.
AB - The task of analysing the inhabitants of the underwater world applies to a wide range of applied problems: construction, fishing, and mining. Currently, this task is applied on an industrial scale by a rigorous review done by human experts in underwater life. In this work, we present a tool that we have created that allows us to significantly reduce the time spent by a person on video analysis. Our technology offsets the painstaking video review task to AI, creating a shortcut that allows experts only to verify the accuracy of the results. To achieve this, we have developed an observation pipeline by dividing the video into frames; assessing their degree of noise and blurriness; performing corrections via resolution increase; analysing the number of animals on each frame; building a report on the content of the video, and displaying the obtained data of the biotope on the map. This dramatically reduces the time spent analysing underwater video data.Also, we considered the task of biotope mass calculation. We correlated the Few-shot learning segmentation model results with point cloud data to achieve that. That provided us with a biotope surface coverage area that allowed us to approximate its volume. Such estimation is helpful for precise area mapping and surveillance.Thus, this paper presents a system that allows detailed underwater biotope mapping using automatic processing of a single camera underwater video data. To achieve this, we combine into a single pipeline a set of deep neural networks that work in tandem.
UR - https://indico.jinr.ru/event/3084/sessions/1751/#20220707
UR - https://pos.sissa.it/429/#session-4601
M3 - Conference contribution
BT - The 6th International Workshop on Deep Learning in Computational Physics (DLCP2022)
Y2 - 6 July 2022 through 8 July 2022
ER -
ID: 102520152