Estimates of Maize Plant Density from UAV RGB Images Using Faster-RCNN Detection Model: Impact of the Spatial Resolution

Abstract

Early-stage plant density is an essential trait that determines the fate of a genotype under given environmental conditions and management practices. The use of RGB images taken from UAVs may replace the traditional visual counting in fields with improved throughput, accuracy, and access to plant localization. However, high-resolution images are required to detect the small plants present at the early stages. This study explores the impact of image ground sampling distance (GSD) on the performances of maize plant detection at three-to-five leaves stage using Faster-RCNN object detection algorithm. Data collected at high resolution (GSD≈0.3 cm) over six contrasted sites were used for model training. Two additional sites with images acquired both at high and low (GSD≈0.6 cm) resolutions were used to evaluate the model performances. Results show that Faster-RCNN achieved very good plant detection and counting (rRMSE=0.08) performances when native high-resolution images are used both for training and validation. Similarly, good performances were observed (rRMSE=0.11) when the model is trained over synthetic low-resolution images obtained by downsampling the native training high-resolution images and applied to the synthetic low-resolution validation images. Conversely, poor performances are obtained when the model is trained on a given spatial resolution and applied to another spatial resolution. Training on a mix of high- and low-resolution images allows to get very good performances on the native high-resolution (rRMSE=0.06) and synthetic low-resolution (rRMSE=0.10) images. However, very low performances are still observed over the native low-resolution images (rRMSE=0.48), mainly due to the poor quality of the native low-resolution images. Finally, an advanced super resolution method based on GAN (generative adversarial network) that introduces additional textural information derived from the native high-resolution images was applied to the native low-resolution validation images. Results show some significant improvement (rRMSE=0.22) compared to bicubic upsampling approach, while still far below the performances achieved over the native high-resolution images.

Comparison of Deep Learning Methods for Detecting and Counting Sorghum Heads in UAV Imagery

“…Therefore, the optimal IoU threshold was 0.3, indicating that in practical target detection studies, appropriate IoU thresholds need to be set according to the size of different detection targets. The sorghum heads in this study are small targets or have a high overlap ratio, so a smaller IoU threshold needed to be selected, and this conclusion was consistent with the results of Velumani et al for corn planting density [69]. On the contrary, for medium and large targets or targets with a low overlap ratio, an IoU threshold of 0.5 or more was selected to obtain better detection results; for example, the average IoU detected by Malambo et al exceeded 0.8 [71] and the IoU value of Tian et al for the detection of apples exceeded 0.85 [49].…”

“…On the other hand, the NMS also removed images in the high overlap region with a high overlap ratio with the maximum confidence prediction frame [68] and the impact of this part was mainly controlled by setting the threshold value of the overlap ratios. Unlike previous studies where the default confidence threshold of 0.5 was commonly used [20,53,69], this study concluded that a confidence of 0.3 was the best for sorghum head recognition. The main reason was that the sorghum heads in this study were weak, small targets on the image and the confidence returned by the prediction frame was generally not high due to factors such as the UAV image, the sorghum heads themselves, and the environmental background (see Section 5.3 for details).…”

Artificial intelligence for automated detection of large mammals creates path to upscale drone surveys

“…Additionally, access to standardized platforms and sensors for long term and large-scale studies is a challenge, which could be overcome with algorithms like ours, potentially independent of ground sampling distances. Some successful examples of this approach are present in the literature for the detection of conifers 66 , crops 67 , and large mammals 15 . To achieve that, inter alia , we needed to assess what are the limits for algorithms to be trained with a range of ground sampling distances, able to accurately classify targets; in addition to evaluations under different whether conditions, backgrounds, and species.…”

High-throughput field crop phenotyping: current status and challenges

“…Similarly, the senescence or stay-green of wheat, maize, and sorghum has been evaluated by UAS RGB or multispectral images ( Hassan et al 2018 , Liedtke et al 2020 , Makanza et al 2018 ). Using UAS-RGB images, the emergence of wheat, rice, maize, and potato was evaluated ( Li et al 2019 , Liu et al 2017 , Velumani et al 2021 , Wu et al 2019 ). In a unique study, Bruce et al (2021) assessed the variation of soybean pubescence using UAS multispectral images.…”