A Novel Wireless Leaf Area Index Sensor Based on a Combined U-Net Deep Learning Model

Leaf area index (LAI) is an important parameter for forestry vegetation canopy structure investigation and ecological environment model study. Traditional ground direct measuring method too time labor consuming, while the remote sensing technique lacks of adequate validation comparative analysis. Here, a novel wireless LAI sensor based on lightweight deep learning (LAINET) has been designed with Raspberry Pi microcomputer LoRa transceiver. The mainly metering pattern system digital hemispherical photo-graphy (DHP) methodology Beer-Lambert law: firstly, crown canopy’s image captured segmented by LAINET, then gap fraction can be extracted to calculate value. Our proposed LAINET consists convolutional neural network (CNN) generative adversarial (GAN). average accuracy semantic segmentation (i.e. CNN part) could reach 0.978, combination GAN super-resolution reconstruction improve measurement more 5.5%. In addition, effectively solves problem low brought environmental effects, separation in sunlight or clear weather improved significantly. So ultimate value calculated precisely stably. Experiment results show that obtains fine error less than 4% when comparing commercial plant analyzer HM-G20. Combined Uninterruptible Power Supply module 5200 mAh, work about 8 months, principally meeting deployment criteria LAI. Therefore, presented this paper great application prospect.