/pdf/activation-functions-in-deep-learning-a-comprehensive-survey-11svx8wh.pdf

Activation functions in deep learning: A comprehensive survey and benchmark

Using a depth camera for crop row detection and mapping for under-canopy navigation of agricultural robotic vehicle

Over the past decade, blockchain technology has attracted a huge attention from both industry and academia because it can be integrated with a large number of everyday applications of modern information and communication technologies (ICT). Peer-to-peer (P2P) architecture of blockchain enhances these applications by providing strong security and trust-oriented guarantees, such as immutability, verifiability, and decentralization. Despite these incredible features that blockchain technology brings to these ICT applications, recent research has indicated that the strong guarantees are not sufficient enough and blockchain networks may still be prone to various security, privacy, and reliability issues. In order to overcome these issues, it is important to identify the anomalous behaviour within the actionable time frame. In this article, we provide an in-depth survey regarding integration of anomaly detection models in blockchain technology. For this, we first discuss how anomaly detection can aid in ensuring security of blockchain based applications. Then, we demonstrate certain fundamental evaluation metrics and key requirements that can play a critical role while developing anomaly detection models for blockchain. Afterwards, we present a thorough survey of various anomaly detection models from the perspective of each layer of blockchain. Finally, we conclude the article by highlighting certain important challenges alongside discussing how they can serve as future research directions for new researchers in the field. 

/pdf/anomaly-detection-in-blockchain-networks-a-comprehensive-13mrw0xf.pdf

Anomaly Detection in Blockchain Networks: A Comprehensive Survey

Currently, deep learning-based visual inspection has been highly successful with the help of supervised learning methods. However, in real industrial scenarios, the scarcity of defect samples, the cost of annotation, and the lack of $a$ priori knowledge of defects may render supervised-based methods ineffective. In recent years, unsupervised anomaly localization (AL) algorithms have become more widely used in industrial inspection tasks. This article aims to help researchers in this field by comprehensively surveying recent achievements in unsupervised AL in industrial images using deep learning. The survey reviews more than 120 significant publications covering different aspects of AL, mainly covering various concepts, challenges, taxonomies, benchmark datasets, and quantitative performance comparisons of the methods reviewed. In reviewing the achievements to date, this article provides detailed predictions and analysis of several future research directions. This review provides detailed technical information for researchers interested in industrial AL and who wish to apply it to the localization of anomalies in other fields.

Deep Learning for Unsupervised Anomaly Localization in Industrial Images: A Survey

Key technologies of machine vision for weeding robots: A review and benchmark

GAN-based anomaly detection: A review

Crop recognition is one of the key processes for robotic weeding in precision agriculture, which remains an open problem due to the unstructured field environment and the wide variety of plant species. It becomes especially challenging when the weeds are prominent and overlap with the crop plants. This paper presents a novel method for recognizing crop plants of field images with a high weed presence. This method segments crop plants from overlapped weeds based on the visual attention mechanism of the human visual system using a convolutional neural network. The network utilizes ResNet-10 as backbone, while introducing side outputs and short connections for multi-scale feature fusion. The Adaptive Affinity Fields method is adopted to improve the segmentation at object boundaries and for fine structures. To train and test the network, a field image dataset has been created which consists of 788 color images with manually segmented annotations. The images are captured under challenging conditions with extremely high weed pressure. The experimental results show that the proposed method can accurately segment crops from weeds and soil, with mean absolute errors less than 0.005 and F-measure scores exceeding 97%. In terms of efficiency, the proposed method can process up to 169 images per second when accelerated by a NVIDIA RTX 2080Ti graphics processing unit (GPU), and operate at approximately 5.6 Hz in a Jetson TX2 embedded computer. The results indicate that the proposed method has the potential to provide an efficient solution for recognizing crop plants, even in the presence of severe weed growth. The code and the dataset are available at https://github.com/ZhangXG001/Real-Time-Crop-Recognition .

/pdf/real-time-crop-recognition-in-transplanted-fields-with-2twtowy7rf.pdf

Real-Time Crop Recognition in Transplanted Fields With Prominent Weed Growth: A Visual-Attention-Based Approach

Controlling weeds with reduced reliance on herbicides is one of the main challenges to move toward a more sustainable agriculture. Robotic weeding is a thought to be a viable way to reduce the environmental loading of agrochemicals while keeping the operation efficiency high. One of the key technologies for performing robotic weeding is automatic detection of crops and weeds in fields. This paper presents an overview on various methods for detecting plants based on machine vision, mainly concentrating on two main challenges: dealing with changing light and crop/weed discrimination. To overcome the first challenge, both physical and algorithmic methods have been proposed. Physical methods can result in a more cumbersome machine while algorithmic methods are less robust. For crop/weed discrimination, deep-learning-based methods have shown obvious advantages over traditional methods based on hand-crafted features. However, traditional methods still hold some merits that can be leveraged to deep-learning-based methods. With the fast development of hardware technologies, researchers should take full advantage of advanced hardware to ease the algorithm design. In the future, the identification of crops and weeds can be more accurate and fine-grained with the support of online databases and computing resources based on the advances in artificial intelligence and communication technologies.

Review of Machine-Vision-Based Plant Detection Technologies for Robotic Weeding

Robotic mechanical weed control is a promising solution to reduce the usage of herbicides. Efficient and accurate detection of crop stems is the premise of most robotic mechanical weeding machines. This paper proposes a unified convolutional neural network model, called UniStemNet, for real-time crop recognition and stem detection. The UniStemNet consists of a backbone network and two subnets to perform the two tasks simultaneously. According to the difference of targets in the two tasks, the varied-span feature fusion structure is established in the subnets. To improve the stem detection performance, a cross-task feature fusion strategy is devised which introduces a top-down guidance from the crop recognition subnet to the stem detection subnet. Experimental results demonstrate that the proposed UniStemNet can significantly outperform the state-of-the-art crop stem detection method, and perform comparably with leadingedge crop recognition methods. The results also validate the remarkable effect of the cross-task feature fusion strategy on improving the stem detection performance. The UniStemNet can process a 400×300 image within 6 ms. The code and dataset are available at https://github.com/ZhangXG001/Real-Time-Crop-Recognition-and-Stem-Localization.git.

A Unified Model for Real-Time Crop Recognition and Stem Localization Exploiting Cross-Task Feature Fusion

Computer-vision-based rust detection is a promising way to enhance the inspection efficiency and accuracy for the steel components of infrastructures. Existing automatic rust detection methods may fail to differentiate and extract different types of rust corrosion, and may require an enormous amount of precisely annotated training data. To resolve these issues, this paper proposes a novel rust segmentation approach based on the Gaussian mixture model (GMM) and superpixel segmentation. The GMM model learns the feature distribution of different degrees of rustiness in the HSV color space and predicts the rustiness probability of each pixel in the input image. The SLIC algorithm is exploited for superpixel segmentation on the input images. Rust/non-rust classification is conducted on the superpixels according to the mean of rustiness probabilities in each superpixel with a threshold generated by the OTSU algorithm. Finally, rust segmentation masks are produced according to the superpixel classification results. The proposed rust segmentation approach simplifies the preprocessing phase and takes into consideration the characteristics of rust areas with different rustiness degrees and the correlation of adjacent pixels. Experimental results show that the proposed approach is robust to the variation of rust features and can generate consistent segmentation results at the boundaries of rusted areas, while only requiring a relatively small training data set. The average processing time for an 120×88 image is 24.283 ms, which indicates that the proposed approach is promising to achieve real-time processing in real-world applications. Our code is available at: https://github.com/lyzx2001/GMM-SLIC-RustDetction

Xiaoguang Zhang

Papers

GAN-based anomaly detection: A review

Real-Time Crop Recognition in Transplanted Fields With Prominent Weed Growth: A Visual-Attention-Based Approach

Review of Machine-Vision-Based Plant Detection Technologies for Robotic Weeding

A Unified Model for Real-Time Crop Recognition and Stem Localization Exploiting Cross-Task Feature Fusion

Automatic Rust Segmentation Using Gaussian Mixture Model and Superpixel Segmentation