Abstract:The changes in the area sequence of eutrophic areas in abnormal water bodies and the anomalies in the water body mask map make it difficult to output the temporal characteristics of water quality parameters, which cannot guide the telemetry process of abnormal water bodies and result in a lower Kappa coefficient for the telemetry results. Therefore, a high-resolution bidirectional LSTM telemetry system for water anomalies is proposed. In terms of hardware, separate designs are made for remote sensing image collectors, telemetry instruments, and power circuits. In terms of software, using piecewise linear regression equations to perform geometric correction on high-resolution remote sensing spectral images of water bodies, solving the problem of image deformation. Based on the corrected remote sensing images, perform morphological reconstruction operations and region growing segmentation to extract water body regions from the images. By using a bidirectional long short-term memory network, an inversion model with strong nonlinear mapping ability and adaptive learning ability is constructed. Spectral data and historical water quality parameters contained in remote sensing images of water bodies are imported into the model, fully utilizing the temporal characteristics of water quality parameters to output current water quality parameters and guide remote sensing of water body anomalies. Using the Isolation Forest algorithm to identify abnormal water quality parameters and derive high-resolution telemetry results for water anomalies. The test results show that the Kappa coefficient of the telemetry results for water anomalies provided by the system exceeds 0.9, meeting the high-resolution monitoring requirements for water environment anomalies.
