Edge Detection Problem Research Articles

STEdge: Self-Training Edge Detection With Multilayer Teaching and Regularization.

Learning-based edge detection has hereunto been strongly supervised with pixel-wise annotations which are tedious to obtain manually. We study the problem of self-training edge detection, leveraging the untapped wealth of large-scale unlabeled image datasets. We design a self-supervised framework with multilayer regularization and self-teaching. In particular, we impose a consistency regularization which enforces the outputs from each of the multiple layers to be consistent for the input image and its perturbed counterpart. We adopt L0-smoothing as the "perturbation" to encourage edge prediction lying on salient boundaries following the cluster assumption in self-supervised learning. Meanwhile, the network is trained with multilayer supervision by pseudo labels which are initialized with Canny edges and then iteratively refined by the network as the training proceeds. The regularization and self-teaching together attain a good balance of precision and recall, leading to a significant performance boost over supervised methods, with lightweight refinement on the target dataset. Through extensive experiments, our method demonstrates strong cross-dataset generality and can improve the original performance of edge detectors after self-training and fine-tuning.

Retroactive correction for white-light dispersion as an edge-detection problem in ultrafast spectroscopies

Retroactive correction for white-light dispersion as an edge-detection problem in ultrafast spectroscopies

Ant Colony Optimization Based Edge Detection Algorithm

The problem of edge detection represents one of the most elementary assignments in image processing, providing an essentialbase for all further study and interpretation of the visual data analysis. This paper proposed an enhanced version of the Ant Colony Optimization (ACO) algorithm for edge detection. The following paper tries to compare the Proposed ACO method with the conventional techniques of edge detection like Canny, Prewitt, and Sobel, using various quantitative metrics like Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Entropy, Natural Image Quality Evaluator (NIQE), and Blind/Referenceless Image Spatial Quality Evaluator (BRISQUE) applied over different images. The datasets for this evaluation are considered as a standard Cameraman, a biological cell, and MRIimage, with and without noise, considering the ranges of complexities and textures.The results of our study prove the competencies of ACO's algorithm. In some cases, it stands out against standard algorithms for MSE and PSNR values and maintains high Entropy values, suggesting the robustness of detail-keeping in an image. Further, the quality assessment of the images by using NIQE and BRISQUE shows the ability of ACO to maintain a natural appearance post-edge detection. In this regard, the studyhighlights that the proposed ACO isan effective method for edge detection in varying image conditions,and, in doing so, it even validates the effectiveness of bio-inspired algorithms in image processing domains.

Boosting edge detection via Fusing Spatial and Frequency Domains

Deep learning-based edge detection methods have shown great advantages and obtained promising performance. However, most of the current methods only extract features from the spatial (RGB) domain for edge detection and the information that can be mined is limited. As a result, they could not work well for the scenarios where the object is similar in color to the background. To combat this challenge, we propose a novel edge detection method by incorporating the features of both spatial and frequency domains, named Fusing Spatial and Frequency Domains (FSFD). A Frequency Perception (FP) module is constructed to extract the edges of objects in the frequency domain, which can avoid the indistinguishable situation in the spatial domain due to similar colors. A Multi-Scale Enhancement (MSE) module is designed to learn multi-scale feature in spatial domain, enabling the model to perceive edges of small objects. Spatial-Frequency Fusion (S2F) module is further introduced to fuse the features of spatial and frequency domains using an online learning manner. Adequate experiments are conducted on popular BSDS500, NYUDV2, and Multicue datasets. The results show that our method can outperform other methods as the state-of-the-art when dealing with the problem of edge detection. The codes will be released on https://github.com/JingDongdong/FSFD.

An Intelligent Apple Identification Method via the Collaboration of YOLOv5 Algorithm and Fast-Guided Filter Theory

Apple-picking robot can promote the development of smart agriculture, and accurate object recognition in complex natural environments using deep learning algorithms is critical. However, research has shown that changes in illumination and object occlusion remain significant challenges for recognition. In order to improve the accuracy of apple apple-picking robot’s identification and positioning of apples in natural environment, a method using YOLOv5 (You Only Look Once, YOLO) combined with fast-guided filter is proposed. By introducing a fast-guided filtering module, the ability to extract image features is improved, and the problem of inaccurate occlusion targets and edge detection is solved; [Formula: see text]-means clustering algorithm is introduced in improving YOLOv5, which can realize automatic adjustment of image size and step size; BiFPN structure is introduced in Neck network to add weighted feature fusion to highlight the detailed features. The results show that the algorithm proposed in this paper can well remove noise information such as occlusion edge blurring in apple images in a natural light environment. In the real orchard environment, the apple recognition accuracy rate reached 97.8%, the recall rate was 97.3% and the recognition rate was about 26.84[Formula: see text]fps. The results show that this research based on YOLOv5 and fast-guided filtering can realize fast and accurate identification of apple fruits in natural environment, and meet the practical application requirements of real-time target detection.

RCF-ST: RICHER CONVOLUTIONAL FEATURES NETWORK WITH STRUCTURAL TUNING FOR THE EDGE DETECTION ON NATURAL IMAGES

Context. The problem of automating of the edge detection on natural images in intelligent systems is considered. The subject of the research is the deep learning convolutional neural networks for edge detection on natural images.
 Objective. The objective of the research is to improve the edge detection performance of natural images by structural tuning the richer convolutional features network architecture.
 Method. In general, the edge detection performance is influenced by a neural network architecture. To automate the design of the network structure in the paper a structural tuning of a neural network is applied. Computational costs of a structural tuning are incomparably less compared with neural architecture search, but a higher qualification of the researcher is required, and the resulting solution will be suboptimal. In this research it is successively applied first a destructive approach and then a constructive approach to structural tuning of the based architecture of the RCF neural network. The constructive approach starts with a simple architecture network. Hidden layers, nodes, and connections are added to expand the network. The destructive approach starts with a complex architecture network. Hidden layers, nodes, and connections are then deleted to contract the network. The structural tuning of the richer convolutional features network includes: (1) reducing the number of convolutional layers; (2) reducing the number of convolutions in convolutional layers; (3) removing at each stage the sigmoid activation function with subsequent calculation of the loss function; (4) addition of the batch normalization layers after convolutional layers; (5) including the ReLU activation functions after the added batch normalization layers. The obtained neural network is named RCF-ST. The initial color images were scaled to the specified size and then inputted in the neural network. The advisability of each of the proposed stages of network structural tuning was reseached by estimating the edge detection performance using the confusion matrix elements and Figure of Merit. The advisability of a structural tuning of the neural network as a whole was estimated by comparing it with methods known from the literature using the Optimal Dataset Scale and Optimal Image Scale.
 Results. The proposed convolutional neural network has been implemented in software and researched for solving the problem of edge detection on natural images. The structural tuning technique may be used for informed design of the neural network architectures for other artificial intelligence problems.
 Conclusions. The obtained RCF-ST network allows to improve the performance of edge detection on natural images. RCF-ST network is characterized by a significantly fewer parameters compared to the RCF network, which makes it possible to reduce the resource consumption of the network. Besides, RCF-ST network ensures the enhancing of the robustness of edge detection on texture background.

The Evaluation of Entropy-based Algorithm towards the Production of Closed-Loop Edge

This research concerns the common problem of edge detection that produces a disjointed and incomplete edge, leading to the misdetection of visual objects. The entropy-based algorithm can potentially solve this problem by classifying the pixel belonging to which objects in an image. Hence, the paper aims to evaluate the performance of entropy-based algorithm to produce the closed-loop edge representing the formation of object boundary. The research utilizes the concept of Entropy to sense the uncertainty of pixel membership to the existing objects to classify pixels as the edge or object. Six entropy-based algorithms are evaluated, i.e., the optimum Entropy based on Shannon formula, the optimum of relative-entropy based on Kullback-Leibler divergence, the maximum of optimum entropy neighbor, the minimum of optimum relative-entropy neighbor, the thinning of optimum entropy neighbor, and the thinning of optimum relative-entropy neighbor. The experiment is held to compare the developed algorithms against Canny as a benchmark by employing five performance parameters, i.e., the average number of detected objects, the average number of detected edge pixels, the average size of detected objects, the ratio of the number of edge pixel per object, and the average of ten biggest sizes. The experiment shows that the entropy-based algorithms significantly improve the production of closed-loop edges, and the optimum of relative-entropy neighbor based on Kullback-Leibler divergence becomes the most desired approach among others due to the production of more considerable closed-loop edge in the average. This finding suggests that the entropy-based algorithm is the best choice for edge-based segmentation. The effectiveness of Entropy in the segmentation task is addressed for further research.

The Evaluation of Entropy-based Algorithm towards the Production of Closed-Loop Edge

This research concerns the common problem of edge detection that produces a disjointed and incomplete edge, leading to the misdetection of visual objects. The entropy-based algorithm can potentially solve this problem by classifying the pixel belonging to which objects in an image. Hence, the paper aims to evaluate the performance of entropy-based algorithm to produce the closed-loop edge representing the formation of object boundary. The research utilizes the concept of Entropy to sense the uncertainty of pixel membership to the existing objects to classify pixels as the edge or object. Six entropy-based algorithms are evaluated, i.e., the optimum Entropy based on Shannon formula, the optimum of relative-entropy based on Kullback-Leibler divergence, the maximum of optimum entropy neighbor, the minimum of optimum relative-entropy neighbor, the thinning of optimum entropy neighbor, and the thinning of optimum relative-entropy neighbor. The experiment is held to compare the developed algorithms against Canny as a benchmark by employing five performance parameters, i.e., the average number of detected objects, the average number of detected edge pixels, the average size of detected objects, the ratio of the number of edge pixel per object, and the average of ten biggest sizes. The experiment shows that the entropy-based algorithms significantly improve the production of closed-loop edges, and the optimum of relative-entropy neighbor based on Kullback-Leibler divergence becomes the most desired approach among others due to the production of more considerable closed-loop edge in the average. This finding suggests that the entropy-based algorithm is the best choice for edge-based segmentation. The effectiveness of Entropy in the segmentation task is addressed for further research.

The Evaluation of Entropy-based Algorithm towards the Production of Closed-Loop Edge

This research concerns the common problem of edge detection that produces a disjointed and incomplete edge, leading to the misdetection of visual objects. The entropy-based algorithm can potentially solve this problem by classifying the pixel belonging to which objects in an image. Hence, the paper aims to evaluate the performance of entropy-based algorithm to produce the closed-loop edge representing the formation of object boundary. The research utilizes the concept of Entropy to sense the uncertainty of pixel membership to the existing objects to classify pixels as the edge or object. Six entropy-based algorithms are evaluated, i.e., the optimum Entropy based on Shannon formula, the optimum of relative-entropy based on Kullback-Leibler divergence, the maximum of optimum entropy neighbor, the minimum of optimum relative-entropy neighbor, the thinning of optimum entropy neighbor, and the thinning of optimum relative-entropy neighbor. The experiment is held to compare the developed algorithms against Canny as a benchmark by employing five performance parameters, i.e., the average number of detected objects, the average number of detected edge pixels, the average size of detected objects, the ratio of the number of edge pixel per object, and the average of ten biggest sizes. The experiment shows that the entropy-based algorithms significantly improve the production of closed-loop edges, and the optimum of relative-entropy neighbor based on Kullback-Leibler divergence becomes the most desired approach among others due to the production of more considerable closed-loop edge in the average. This finding suggests that the entropy-based algorithm is the best choice for edge-based segmentation. The effectiveness of Entropy in the segmentation task is addressed for further research.

Enhancement of Potential Field Source Boundaries Using the Hyperbolic Domain (Gudermannian Function)

Horizontal boundary identification of causative sources is an essential tool in potential field data interpretation due to the feasibility of automatically retrieving the boundary information of subsurface gravity or geomagnetic structures. Although many approaches have been proposed to address these issues, it is still a hot research topic for many researchers to derive novel methods or enhance existing techniques. We present two high-resolution edge detectors based on the Gudermannian function and the modifications of the second-order derivative of the field. The effectiveness of the newly proposed filters was initially tested on synthetic gravity anomalies and geomagnetic responses with different assumptions (2-D and 3-D; imposed and superimposed; noise-free and noise-contaminated). The obtained results verified that the two novel methods yield the capability of producing high-resolution, balanced amplitudes and accurate results for better imaging causative sources with different geometrical and geophysical properties, compared with the other nine representative edge enhancement techniques. Furthermore, the yielded results from the application of the two strategies to a real-world aeromagnetic data set measured from the Central Puget Lowland (C.P.L) of the United States and a gravity data set surveyed from the Jalal Abad area of Kerman province, Iran, with detailed comparative studies validated that the edges identified via the two methods are in good agreement with the major geological structures within the study areas and the determined lateral information using the tilt-depth, top-depth estimation method. These features make them valuable tools for solving edge detection problems.

Linear operators with vector masks in digital image processing problems

The paper shows that it is expedient to use vector masks for solving some types of digital image processing problems. The main advantage of vector masks compared to matrix masks is that they reduce the computational complexity of algorithms while maintaining, and in some problems even improving, quality indicators. The article demonstrates examples of the use of vector masks in the problem of estimating the level of discrete white noise in an image, forming a basis for constructing a correctly working sigma filter, which are used for obtaining smoothed partial derivative estimates in the problem of edge detection and detecting straight lines in a contour image. The work uses results obtained by the authors in their earlier publications.

DFTNet: Dual Flow Transformer Network for Conveyor Belt Edge Detection

In traditional conveyor belt edge detection methods, contact detection methods have a high cost. At the same time noncontact detection methods have low precision, and the methods based on the convolutional neural network are limited by the local operation features of the convolution operation itself, causing problems such as insufficient perception of long-distance and global information. In order to solve the above problems, a dual flow transformer network (DFTNet) integrating global and local information is proposed for belt edge detection. DFTNet could improve belt edge detection accuracy and suppress the interference of belt image noise. In this paper, the authors have merged the advantages of the traditional convolutional neural network’s ability to extract local features and the transformer structure’s ability to perceive global and long-distance information. Here, the fusion block is designed as a dual flow encoder–decoder structure, which could better integrate global context information and avoid the disadvantages of a transformer structure pretrained on large datasets. Besides, the structure of the fusion block is designed to be flexible and adjustable. After sufficient experiments on the conveyor belt dataset, the comparative results show that DFTNet can effectively balance accuracy and efficiency and has the best overall performance on belt edge detection tasks, outperforming full convolution methods. The processing image frame rate reaches 53.07 fps, which can meet the real-time requirements of the industry. At the same time, DFTNet can deal with belt edge detection problems in various scenarios, which gives it great practical value.

An Image Edge Detection Algorithm Based on an Artificial Plant Community

Image edge detection is a difficult task, because it requires the accurate removal of irrelevant pixels, while retaining important pixels that describe the image’s structural properties. Here, an artificial plant community algorithm is proposed to aid in the solving of the image edge detection problem. First, the image edge detection problem is modeled as an objective function of an artificial plant community searching for water sources and nutrients. After many iterations, the artificial plant community is concentrated in habitable areas that are rich in water sources and nutrients, that is, the image edges, and the nonhabitable zones that are not suitable for living are deserted, that is, the nonedges. Second, an artificial plant community algorithm is designed to solve the objective function by simulating the growth process of a true plant community. The living behavior of the artificial plant community includes three operations: seeding, growing, and fruiting. The individuals in the plant community also correspond to three forms, namely seeds, individuals, and fruit. There are three fitness comparisons in each iteration. The first fitness comparison of each iteration is carried out during the seeding operation. Only the fruit with higher fitness levels in the last iteration can become seeds, while the fruit with low fitness levels die, and some new seeds are randomly generated. The second fitness comparison is implemented in the growing operation. Only the seeds with higher fitness levels can become individuals, but the seeds with lower fitness levels will die; thus, the community size will decrease. The third fitness comparison is in the fruiting operation, where the individual with the greatest fitness can produce an identical fruit through parthenogenesis, and the individuals with higher fitness levels can learn from each other and produce more fruit, so the population size can be restored. Through the continuous cycle of these three operations, the artificial plant community will finally determine the edge pixels and delete the nonedge pixels. Third, the experiment results reveal how the proposed algorithm generates the edge image, and the comparative results demonstrate that the proposed artificial plant community algorithm can effectively solve the image edge detection problems. Finally, this study and some limitations are summarized, and future directions are suggested. The proposed algorithm is expected to act as a new research tool for solving various complex problems.

MTF Measurement by Slanted-Edge Method Based on Improved Zernike Moments.

The modulation transfer function (MTF) is an important parameter for performance evaluation of optical imaging systems in photogrammetry and remote sensing; the slanted-edge method is one of the main methods for measuring MTF. To solve the problem of inaccurate edge detection by traditional methods under the conditions of noise and blur, this paper proposes a new method of MTF measurement with a slanted-edge method based on improved Zernike moments, which firstly introduces the Otsu algorithm to automatically determine the Zernike moment threshold for sub-pixel edge detection to precisely locate the edge points, then obtains LSF through edge point projection, ESF sampling point acquisition, smoothing, fitting, taking ESF curve differential and Gaussian fitting, and finally, accurately obtaining MTF by LSF Fourier transform and modulo normalization. Based on simulation experiments and outdoor target experiments, the reliability of the proposed algorithm is verified by the deviations of slanted-edge angle and MTF measurement, and the tolerance degree of edge detection to noise and ambiguity are analyzed. The results show that compared with ISO 12233, OMNI-sine method, Hough transform method and LSD method, this algorithm has the highest edge detection accuracy, the maximum tolerance of noise and ambiguity, and also improves the accuracy of MTF measurement.

Efficient Object Detection in SAR Images Based on Computation-Aware Neural Architecture Search

Remote sensing techniques are becoming more sophisticated as radar imaging techniques mature. Synthetic aperture radar (SAR) can now provide high-resolution images for day-and-night earth observation. Detecting objects in SAR images is increasingly playing a significant role in a series of applications. In this paper, we address an edge detection problem that applies to scenarios with ship-like objects, where the detection accuracy and efficiency must be considered together. The key to ship detection lies in feature extraction. To efficiently extract features, many existing studies have proposed lightweight neural networks by pruning well-known models in the computer vision field. We found that although different baseline models have been tailored, a large amount of computation is still required. In order to achieve a lighter neural network-based ship detector, we propose Darts_Tiny, a novel differentiable neural architecture search model, to design dedicated convolutional neural networks automatically. Darts_Tiny is customized from Darts. It prunes superfluous operations to simplify the search model and adopts a computation-aware search process to enhance the detection efficiency. The computation-aware search process not only integrates a scheme cutting down the number of channels on purpose but also adopts a synthetic loss function combining the cross-entropy loss and the amount of computation. Comprehensive experiments are conducted to evaluate Darts_Tiny on two open datasets, HRSID and SSDD. Experimental results demonstrate that our neural networks win by at least an order of magnitude in terms of model complexity compared with SOTA lightweight models. A representative model obtained from Darts_Tiny (158 KB model volume, 28 K parameters and 0.58 G computations) yields a faster detection speed such that more than 750 frames per second (800×800 SAR images) could be achieved when testing on a platform equipped with an Nvidia Tesla V100 and an Intel Xeon Platinum 8260. The lightweight neural networks generated by Darts_Tiny are still competitive in detection accuracy: the F1 score can still reach more than 83 and 90, respectively, on HRSID and SSDD.

A robust edge detection technique based on Matching Pursuit algorithm for natural and medical images

Edge detection has been used widely in image processing, image recognition, and computer vision. Several edge detection techniques have been suggested over time; however, each has a lot of disadvantages. Our method has some steps. At first, the image should be converted into a grayscale image. A 3*3pattern grid is used for producing a discrete vector for any pixel in the present study. Secondly, the length of the pixel vector is increased to linearize and expand. Thirdly, the Matching Pursuit algorithm is applied to the image, and the residual part of the Matching Pursuit algorithm is extracted. Finally, two Hysteresis thresholds have been used for detecting edge pixels. The edge detection problem has been converted into a signal processing problem, so it will be less time-consuming. Moreover, the proposed algorithm is insensitive to noise, and this method can detect weak edge pixels. Also, the performance of the proposed algorithm is compared with three known edge detection algorithms, SOBEL, Laplacian, Canny, and CASENet and the results show that we have a high value of performance parameters such as PR, PSNR, and FOM compared to other algorithms. Furthermore, it is very an effective and reliable method to diagnose and treat diseases in medical images.

Critical node/edge detection problems on trees

We consider the problem of removing a limited subset of nodes and/or edges from a graph in order to minimize the so-called pairwise connectivity of the residual graph, which is defined as the total cost of the pairs of nodes still connected by a path. This is a well-studied version of a family of problems known as critical node or edge detection problems. However, while most of the literature focuses on deleting nodes or edges separately, we allow the simultaneous removal of nodes and edges. We consider both the case in which the nodes and edges removed must satisfy a joint weight limit, and the case in which two separate weight limits are given for nodes and edges. We study the complexity of several problems of this type when the given graph is a tree, providing NP-hardness results or polynomial-time algorithms for the different cases that we analyze.

Borehole magnetometry to determine the depth of reinforced foundations in telecommunication towers

Technical information about deep foundations of built structures may in some cases be unavailable or incomplete due to the lack of construction documentation and electronic media in which the reports are stored. This problem is particularly relevant for telecommunication towers constructed two-to-three decades ago but still under operation and subject to new loading demands from expanding networks. Nondestructive evaluation with geophysical methods has been applied to determine the depth to the bottom of deep foundations, a key parameter to evaluate their bearing capacity and settlements. This is done by lowering a borehole magnetometry (BM) sensor down a drillhole installed close to the unknown structure, thereby measuring the magnetic field along the testing drillhole. Determining the bottom depth of the structure using BM data configures an edge detection problem: a classical problem in magnetic data interpretation, but in this case with poor data coverage (only a profile close to the structure) to characterize a complex magnetization distribution. Common practices to locate pile bottom from characteristic peak or inflection points in the observed magnetic anomaly may work at some sites but usually fail to account for the complex anomaly patterns from built structures. We summarize a set of numerical and field-work applications showing that no single characteristic attribute of the observed field can undoubtedly be used to determine the foundation depth. We verify that most of the measured field along the testing drillhole is disturbed by fields from facilities at the ground surface and develop a data inversion procedure enforcing data fitting to measurements at the lower portions of the drillhole. The efficacy of this technique is illustrated with field tests in different geologic scenarios with results compared with as-built documentation and/or independent evaluation with other methods.

Postprocessing of Edge Detection Algorithms with Machine Learning Techniques

In this paper, machine learning (ML) techniques are applied at an early stage of Image Processing (IP). The learning procedures are usually applied from at least the image segmentation level, whereas, in this paper, this is done from a lower processing level: the edge detection level (ED). The main objective is to solve the edge detection problem through ML techniques. The proposed methodology is based on a classification of edges made pixel by pixel, but the predictors employed for the ML task include information about the pixel neighborhood and structures of connected pixels called edge segments. The Sobel operator is employed as input. Making use of 50 images that belong to the Berkeley Computer Vision data set, the average performance of the validation sets when employing our Neural Networks method reached an F-measure significatively higher than with the Sobel operator. The experiment results show that our post-processing technique is a promising new approach for ED.

Multilevel Image Edge Detection Algorithm Based on Visual Perception

Multilevel image edge repair results directly affect the follow-up image quality evaluation and recognition. Current edge detection algorithms have the problem of unclear edge detection. In order to detect more accurate edge contour information, a multilevel image edge detection algorithm based on visual perception is proposed. Firstly, the digital image is processed by double filtering and fuzzy threshold segmentation; Through the analysis of the contour features of the moving image, the threshold of the moving image features is set, and the latest membership function is obtained to complete the multithreshold optimization. Adaptive smoothing is used to process the contour of the object in the moving image, and the geometric center values of the two adjacent contour points within the contour range are calculated. According to the calculation results, the curvature angle is further calculated, and the curvature symbol is obtained. According to the curvature symbol, the contour features of the moving image are detected. The experimental results show that the proposed algorithm can effectively and accurately detect the edge contour of the image and shorten the reconstruction time, and the detection image resolution is high.