An improved Otsu method using the weighted object variance for defect detection

Detection of skin defects on mangoes based on hyperspectral imaging combined with band ratio and improved Otsu method

An improved maximum between-cluster variance (Otsu) method based on iteration was proposed for the purpose of realizing target recognition and target tracking effectively. Improved Otsu method was used to segment image collected by CCD, so as to extract the target from image and finish the target localization. By calculating centroid of the target in a single frame and extracting feature point of outline constituted by centroids of the single-frame target within unit interval, a polygon can be formed, the gravity center of which was taken as the target position. Filter model was established of which the control input is nonzero based on Kalman filter Principle, which was combined with target position at this moment can be used to predict the gravity center of the target at the next moment. The experiment and simulation results indicated that improved Otsu method can solve the problem that target and background are segmented mistakenly when image thresholding. Kalman filtering was able to predict target trajectory effectively after target being recognized. The requirement of target tracking was satisfied and target tracking was realized through this improved Otsu target tracking algorithm based on iteration.

At present there are some worm detection systems, primarily for a single LAN or with hardware router environment, which main use worm propagation characteristics for detection and has high false alarm rate, but it is not applicable for large-scale network for detecting. This paper presents a distributed worm detection technology, which is divided into two parts, client end and the console end program. The system uses the rule-based detection methods to monitor network worms, and the console side manages and coordinates the multiple side work of detecting. Experimental results show that the method can be good applicable for worms conduct surveillance based on a single or multiple local area network and is used for worm alarming, the method has high detection rate and low false alarm rate.

At present there are some worm intrusion detection systems, primarily for a single LAN or with hardware router environment, which are not applicable for large-scale network detection or have high false alarm rate by using only worm propagation characteristics for detection. This paper analyzed worm non-linear propagation models and drew out the worm transmission curves. Then a distributed worm detection technology is designed. The novel distributed worm detection system consists of two parts, client end and console end programs. The system uses rule-based detection method to monitor network worms, and the console side manages and coordinates detection work of the client sides. Experimental results show that the technology is a good solution to worm detection in multiple network environments which can give an alarm with high detection rate and low false alarm rate when the known worm appears.

As a classical research direction in the field of computer vision, anomaly detection has a very wide range of application prospects. Surface defect detection is an extremely important part of the industrial production process. Most of the traditional defect detection solutions are based on machine vision or supervised learning paradigms. These methods require the use of a large number of supervised defect samples, but it is difficult to obtain sufficient defect data in the production process. Due to various random factors, it is not easy to identify all types of defects during the inspection phase. With the development of deep unsupervised learning in recent years, autoencoder or generative adversarial networks are widely used in surface defect detection problem, which rely solely on the modeling of good data to discriminate abnormal defects. Therefore, in this paper, we propose a neural architecture search-based defect detection method. We combine neural architecture search and generative adversarial learning to automate the search of encoder-decoder structures with a gradient optimization-based search method. The method learns a pair of an anomalous image and its anomaly-free reconstruction, while simultaneously learning a decision boundary between normal and anomalous examples. Neural architecture search helps to search for a task-appropriate network architecture with better performance. Finally, the proposed method is evaluated on the industrial inspection dataset MVTec AD, and the effectiveness of the model detection is measured by AUROC metrics, demonstrating the validity of the methods is demonstrated by both quantitative and qualitative analysis.

Aiming at the shortcomings of manual detection methods for surface defects of bearing outer rings, a surface defect detection method based on machine vision is proposed. The common defects on the bearing surface are scratches, cracks, and peeling. The defect images are analyzed by the LabVIEW image processing module. Make fast and accurate inspections. In this paper, a CMOS industrial camera is used to obtain the bearing image, and the gray histogram of the image is analyzed to determine whether it is a defective bearing. In order to solve the problem of uneven illumination in image segmentation, it is proposed to first use morphological processing for background. Then, the difference map containing defect information is obtained, and then the Otsu method is used for image segmentation; an edge detection method is designed based on morphology, which can achieve complete edge extraction and good denoising effect. Experiments show that the method has strong practicability and self-adaptability, and can accurately detect bearing surface defects.

Steel plate surface defects seriously reduce the steel wear resistance,high temperature resistance,corrosion resistance,fatigue resistance and other properties.Therefore,the detection of plate surface defects is very important.This paper proposes a new method to detect steel defects based on machine vision.Collecting images of steel plate surface in various light conditions are discussed.Firstly,the defect images are preprocessed,and then the preprocessed images are changed to binary images and are processed morphologically.Finally,the image background and object graphics are separated,and the surface defect features are extracted to calculate the defect area and perimeter.After the calibration,the defect’s area and length can be obtained.The experimental results show that the method is of reliability and repeatability.

Internal defects detection and classification in hollow cylindrical surfaces using single shot detection and MobileNet

An intelligent wall inspection robot with sensors and embedded image processing system was studied based on a combination of vacuum-suction technique and four-wheel drive method to achieve good balance between strong adhesion force and high mobility. To obtain more stable adhesion forces, the height and weight of the robot body were reduced by analyzing the relation between the principal physical input variables and the output performance, such as the suction force and momentum. An object detection system on the embedded programming system for automatic wall inspection was also developed. For successful detection, threshold methods were compared and an improved Otsu method was adopted so that it could handle both unimodal and bimodal distributions well. Through the improvement, the robot could move upward a wall at a speed of 5.3 m/min and carry a payload of at least 7.5 kg in addition to its self-weight and it could also intelligently detect and classify small objects on the wall.

The advances in digital image processing technology, the increasing availability of computing resources and highdefinition resolution cameras have resulted in a great deal of interest in object detection algorithms in defects inspection of industrial products. This paper proposes a complete framework to cope with the surface defects detection of mobile phone panel workpieces. The proposed method has a set of consecutive handling processes: original image cutting and standardization, image blocks segmentation and feature extraction, machine learning embedded defects detection. We empirically prove the capability of the proposed method for detecting defects of mobile phone panel surface. The comparison with different algorithms further show that our method achieves excellent performance for mobile phone panel surface defects detection task in both detecting accuracy and detecting speed.

The detection of rail surface defects is an important tool to ensure the safe operation of rail transit. Due to the complex diversity of track surface defect features and the small size of the defect area, it is difficult to obtain satisfying detection results by traditional machine vision methods. The existing deep learning-based methods have the problems of large model sizes, excessive parameters, low accuracy and slow speed. Therefore, this paper proposes a new method based on an improved YOLOv4 (You Only Look Once, YOLO) for railway surface defect detection. In this method, MobileNetv3 is used as the backbone network of YOLOv4 to extract image features, and at the same time, deep separable convolution is applied on the PANet layer in YOLOv4, which realizes the lightweight network and real-time detection of the railway surface. The test results show that, compared with YOLOv4, the study can reduce the amount of the parameters by 78.04%, speed up the detection by 10.36 frames per second and decrease the model volume by 78%. Compared with other methods, the proposed method can achieve a higher detection accuracy, making it suitable for the fast and accurate detection of railway surface defects.

In this study, a method of thresholding is proposed that based on the histogram shape of each image, proposes the more appropriate technique from the famous and common thresholding techniques. Thresholding, which is a commonly used operation for image processing, is the selection of one of the image pixels that determines the border for background and foreground of the image. After determining a suitable threshold, the image can be converted to a binary image and from this binary image, which has a very small size, informations can be extracted for utilization in various scientific subjects. From histogram tests of images, examination of criteria extracted from the histogram shape of each image, and empirical knowledge, we designed an expert system that proposes the suitable thresholding technique for an image based on the histogram of that image. We used modeling and matching for designing this system in such way that a model was produced from the histogram of the most suitable result for each thresholding technique and was stored on a knowledge base. Afterwards, a system is coded to receive the input image and after matching this image with the above models, the technique with the most model proximity to the histogram of the input image is proposed as the more appropriate technique from the methods of thresholding. Nowadays thresholding, which is the preprocessing for each image process, is widely used. So far, many techniques and methods of thresholding have been proposed. Although all the techniques are useful, the results vary for each image. Sometimes a certain technique is more appropriate for an image than the rest of the techniques. Based on experimentation and qualitative reasoning which will be mentioned later, we concluded what techniques are more appropriate for different shapes of histograms. The application of the proposed method is to facilitate selection of a suitable thresholding technique for an image. Furthermore, the proposed method was tested on many images and the results showed that by using this method, choosing a suitable thresholding method can be greatly automatized.

The detection of defects on automotive coated surfaces is of paramount importance to ensure the quality of automotive appearance. However, due to the challenges posed by the lack of sufficient samples of product-specific coated defects, the uneven number of species, and the difficulty of detecting defects in small targets due to the complex background of the body surface. The detection of defects on automotive coated surfaces still relies on experienced manual labor to a significant extent. To address these issues, Poisson fusion is employed to enhance the data for generating defect images, thereby mitigating the limitations of insufficient coated defect samples and an imbalanced distribution of defect species. Based on this, a new YOLOv8-SC defect detection model is proposed, in which the C2f-Star is used to replace the original C2f structure. This aims to improve the model’s ability to extract defects of varying sizes in complex backgrounds, while reducing the number of model parameters and the amount of computation. The content-guided attention fusion (CGAFusion) module, situated before each detection head of the model, has been designed to enhance the model’s capacity to extract defects of varying sizes and complexity. This is achieved through the adaptive fusion of low-level and high-level features. The proposed model’s superiority in terms of detection metrics has been demonstrated through the use of example validation. The experiments were conducted on the enterprise dataset CPD-DET and the public dataset NEU-DET. The results demonstrated that the defective image generation and data enhancement method could significantly improve detection performance and have good generalization. The proposed YOLOv8-SC reduces the model parameters by 12.2% compared to the normal model. Additionally, the mean average precision (mAP) at 50% and 50%–95% thresholds are enhanced by 4.7% and 2.3%, respectively. Furthermore, the accuracy of the model exhibits a discernible positive growth trend with an increase in the sample set size. The design of an automotive coated surface defect detection system is presented at last. When deployed in industrial settings, this system can facilitate the intelligent detection of coated defects.

Timely and accurate monitoring of the spatial distribution of wheat is crucial for wheat field management, growth monitoring, yield estimation and prediction. In this study, a multitemporal index, termed the winter wheat mapping index (WWMI), was constructed for automatic winter wheat mapping on the basis of Sentinel-2 enhanced vegetation index (EVI) time series and wheat phenological features. Henan, an important winter wheat production province in China, was selected as the study area. Zhumadian, the primary wheat-growing city in Henan, was the test area. Both empirical and automatic threshold (Otsu) methods were adopted to calculate the optimal threshold of the WWMI. The performance of WWMI in winter wheat mapping was compared at object-oriented and pixel-based levels. The proposed WWMI separated winter wheat and nonwinter wheat areas well, thus achieving highly accurate winter wheat mapping. In Zhumadian, the empirical threshold method performed better than the Otsu method, but the former relied on official statistics to iteratively adjust the WWMI threshold. In Henan, the mapping accuracy achieved by the Otsu method was higher than that achieved by the empirical threshold method, with mean relative errors (MREs) of 6.78% and 19.87% at the municipal and county levels, respectively. This was because, compared with the empirical threshold method, the Otsu method did not rely on official statistics and adaptively determined the optimal threshold of the WWMI for each city in Henan, thus fully considering wheat growth state differences in different cities. In addition, the object-oriented WWMI performed better than the pixel-based WWMI in wheat mapping. The results further demonstrated the feasibility of combining the WWMI with the Otsu method for automatic winter wheat mapping at large extents, which will provide a theoretical basis for identifying other food crops.

This study aims to enhance the accuracy and robustness of surface defect detection in steel strips. Addressing the challenges of complex defect morphology, low detection accuracy, and high false alarm rates in existing methods, this paper introduces deep learning techniques. It optimizes the YOLOv7 algorithm, proposing multiple improvement strategies. Specifically, these include: employing deformable convolutions (DConv) to address morphological variations in defects; designing a bidirectional cross-attention feature fusion module (BCC) to leverage information across different feature layers fully; constructing a decoupled detection head (TSCODE) to optimize classification and localization tasks; utilizing the SIoU loss function to accelerate bounding box regression; and applying the K-means++ algorithm to obtain predefined anchor boxes. Experimental results demonstrate that the optimized algorithm achieves a mean average precision (mAP) of 80.1% on the strip steel surface defect dataset, representing a 6.0% improvement over the original method. The proposed optimization strategies significantly enhance the performance of strip steel surface defect detection.