Investigation of salt stress effects on maize seedling phenotypic traits based on the PointCornNet point cloud segmentation model

To address the limitations of traditional crop phenotyping methods, such as slow data collection, high error susceptibility, and seedling damage, we proposed a non-destructive approach for phenotypic trait detection in maize seedlings to enhance breeding efficiency. We developed an improved point cloud segmentation model, PointCornNet, based on PointNet++, by integrating the CBAM attention mechanism, replacing the original loss function with Varifocal Loss, and incorporating the CronDBSCAN clustering algorithm to enhance segmentation accuracy and enable both semantic and instance segmentation. Comparative experiments confirmed the improved model performance. Phenotypic parameters—including plant height, canopy width, volume, and surface area—were calculated from the segmented point clouds. The coefficient of determination (R²) between calculated and manually measured values for plant height and canopy width reached 0.99 and 0.96, respectively, demonstrating the accuracy of the method and non-destructive nature. Using PointCornNet and the phenotyping algorithm, we measured 3D morphological changes of maize seedlings under different NaCl concentrations during the first six days after sowing. The results showed that salt stress significantly inhibited seedling growth, with stronger inhibition at higher NaCl concentrations. Increased salt concentration delayed initial seedling emergence and led to gradual decreases in plant height, canopy width, volume, surface area, and their respective growth rates.