An Effective Eichhornia Crassipes Growth Rate Detection and Segmentation Using Adaptive swinResUnet++

A new Eichhornia crassipes spread detection mechanism is developed using an advanced deep learning strategy with multi-spectral images to identify the growth rate of the Eichhornia crassipes. Initially, the required multispectral images are acquired and given to the Intelligent 3D-Yolo Network (I3DYoloNet) is used for detecting the Eichhornia Crassipes. This detection finds whether the Eichhornia crassipes are present or not over a particular region. Here, the parameters from the I3DYoloNet are optimized using the developed Position Upgraded Tuna Swarm with Ageist Spider Monkey Optimization (PU-TSASMO). After detecting the Eichhornia crassipes, the detected region is segmented using the Adaptive SwinResUnet++ (ASRUnet++) for finding the growth rate of Eichhornia crassipes, where the parameters are tuned using the same PU-TSASMO. Segmentation yields the precise infested area in each image, and the growth rate is calculated by comparing this area across time-series images (with optional biomass conversion if needed). Here, the PU-TSASMO-ASRUnet++ achieves 96.52% accuracy, and the designed PU-TSASMO-I3DYoloNet attained 96.63%, which is higher than the other existing frameworks like FCM, KMeans, UNet, and swin_ResUNet++ on a Global Water Hyacinth Invasion database. The results obtained from the experiments are compared with the existing Eichhornia crassipes growth rate detection models to ensure the detection performance.