Super Resolution Generative Adversarial Network (SRGANs) for Wheat Stripe Rust Classification

Wheat yellow rust is a common agricultural disease that affects the crop every year across the world. The disease not only negatively impacts the quality of the yield but the quantity as well, which results in adverse impact on economy and food supply. It is highly desired to develop methods for fast and accurate detection of yellow rust in wheat crop; however, high-resolution images are not always available which hinders the ability of trained models in detection tasks. The approach presented in this study harnesses the power of super-resolution generative adversarial networks (SRGAN) for upsampling the images before using them to train deep learning models for the detection of wheat yellow rust. After preprocessing the data for noise removal, SRGANs are used for upsampling the images to increase their resolution which helps convolutional neural network (CNN) in learning high-quality features during training. This study empirically shows that SRGANs can be used effectively to improve the quality of images and produce significantly better results when compared with models trained using low-resolution images. This is evident from the results obtained on upsampled images, i.e., 83% of overall test accuracy, which are substantially better than the overall test accuracy achieved for low-resolution images, i.e., 75%. The proposed approach can be used in other real-world scenarios where images are of low resolution due to the unavailability of high-resolution camera in edge devices.

ERDBNet: Enhanced Residual Dense Block Net --A New Net to Rich ESRGAN Image Details

Super resolution is applied in many digital image fields. In many cases, only a set of low-resolution images can be obtained, but the image needs a higher resolution, and then SR needs to be applied. SR technology has undergone years of development. Among them, SRGAN is the key work to introduce GAN into the SR field, which can truly restore a large number of details on the basis of low-pixel pictures. ESRGAN is a further improvement on SRGAN. By removing the BN layer in SRGAN, the effect of artifacts in SRGAN is eliminated. However, there is still a problem that the restoration of information on small and medium scales is not accurate enough. The proposed ERDBNet improve the model on the basis of ESRGAN, and use the ERDB block to replace the original RRDB block. The new structure uses a three-layer dense block to replace the original dense block, and a residual structure of the starting point is added to each dense block. The pre-trained network can reach a PSNR of 30.425 after 200k iterations, and the minimum floating PSNR is only 30.213. Compared with the original structure, it is more stable and performs better in the detail recovery of many low-pixel images.

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

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.

Robust Multi-Frame Super-Resolution Based on Adaptive Half-Quadratic Function and Local Structure Tensor Weighted BTV

It is difficult to improve image resolution in hardware due to the limitations of technology and too high costs, but most application fields need high resolution images, so super-resolution technology has been produced. This paper mainly uses information redundancy to realize multi-frame super-resolution. In recent years, many researchers have proposed a variety of multi-frame super-resolution methods, but it is very difficult to preserve the image edge and texture details and remove the influence of noise effectively in practical applications. In this paper, a minimum variance method is proposed to select the low resolution images with appropriate quality quickly for super-resolution. The half-quadratic function is used as the loss function to minimize the observation error between the estimated high resolution image and low-resolution images. The function parameter is determined adaptively according to observation errors of each low-resolution image. The combination of a local structure tensor and Bilateral Total Variation (BTV) as image prior knowledge preserves the details of the image and suppresses the noise simultaneously. The experimental results on synthetic data and real data show that our proposed method can better preserve the details of the image than the existing methods.