Machine Vision Techniques Research Articles

Application of Deep Learning-Based Object Detection Techniques in Fish Aquaculture: A Review

Automated monitoring and analysis of fish’s growth status and behaviors can help scientific aquaculture management and reduce severe losses due to diseases or overfeeding. With developments in machine vision and deep learning (DL) techniques, DL-based object detection techniques have been extensively applied in aquaculture with the advantage of simultaneously classifying and localizing fish of interest in images. This study reviews the relevant research status of DL-based object detection techniques in fish counting, body length measurement, and individual behavior analysis in aquaculture. The research status is summarized from two aspects: image and video analysis. Moreover, the relevant technical details of DL-based object detection techniques applied to aquaculture are also summarized, including the dataset, image preprocessing methods, typical DL-based object detection algorithms, and evaluation metrics. Finally, the challenges and potential trends of DL-based object detection techniques in aquaculture are concluded and discussed. The review shows that generic DL-based object detection architectures have played important roles in aquaculture.

Point and interval prediction of the effective length of hot-rolled plates based on IBES-XGBoost

The prediction of the effective length of steel plates is essential for improving the yield. Due to the complexity of variables and the lack of measuring means, there are few efficient and accurate models for the prediction of effective length. In this paper, a novel data-driven model is proposed to predict the effective length of hot rolled plates. The dataset used for the problem is established using machine vision techniques. An improved bald eagle search algorithm is used to optimize extreme gradient boosting (IBES-XGBoost) for prediction. Based on the point prediction results, the effective length interval of plates is estimated by combining the kernel density estimation method (KDE), which effectively reduces the risk caused by errors. The main work was completed in Python programming language and the validation shows that the proposed model can obtain highly accurate point and interval prediction results, which can provide accurate data reference for billet design and plane shape control of plates.

Classification of arecanut using machine learning techniques

In agricultural domain research, image processing and machine learning techniques play an important role. This paper provides a unique solution for classifying the good and defective arecanuts based on their color, texture, and density value. In the market different varieties of arecanut are available. Usually, qualitative sorting is done manually, and this can be replaced by applying machine vision techniques to grade the arecanut. Classification of arecanut based on quality is done using various machine learning techniques and it is observed that artificial neural networks give good results compared to other classifiers like logistic regression, <em>k</em>-nearest neighbor, naive Bayes classifiers, and support vector machine. A unique density feature is considered here for better classification. The result of classifiers without considering the density feature is compared with respect to the density feature and it is observed that artificial neural networks work better than the others. The proposed method works effectively for classifying arecanut with an accuracy of 98.8%.

Traffic load simulation on bridges based on improved non-contact machine vision technique and multi-source information fusion

This study presented a novel traffic load simulation method. Specifically, an improved You Only Look Once (YOLO) detector based on the k-means++ clustering algorithm is implemented to identify vehicle types, and the multi-objective tracking algorithm is used to obtain the spatial location and velocity information of vehicles. Based on the results of detection and tracking, multi-source information fusion techniques are applied to build a vehicle database. Meanwhile, a Monte Carlo-Intelligent Driving Model can be developed for traffic load simulation based on the vehicle database. The applicability and effectiveness of the method are verified by the suspension bridge field test. The results show that the identification accuracy of vehicle type and vehicle number is about 88.28% and 74.09%, respectively. Meanwhile, the vehicle speed error is less than 6.5%. The method can be used to provide a reasonable traffic load model for large-span bridges without the aid of a deck-embedded weigh-in-motion system.

Characterizing Mosquito Biting Behavior Using the BiteOscope.

The biteOscope enables the high-resolution monitoring and video recording of blood-feeding mosquitoes. Mosquito biting is induced by combining host cues, an artificial bloodmeal, a membrane, and a transparent heater in a transparent behavioral arena. Machine vision techniques enable the tracking and pose estimation of individual mosquitoes to discern behavior and resolve individual feeding events. The workflow allows multiple replicates and large amounts of imaging data to be generated rapidly. These data are suitable for downstream analysis using machine learning tools for behavioral analysis, allowing subtle behavioral effects to be characterized.

Nondestructive determination of common indicators of beef for freshness assessment using airflow-three dimensional (3D) machine vision technique and machine learning

Assessment of beef freshness is critical to ensure a safe and sustainable beef supply chain. This work aims to investigate the feasibility of the viscoelasticity of beef to evaluate its freshness using the airflow-3D machine vision technique and machine learning models. The 3D camera was continuously used to acquire deformation images under the action of the airflow. The obtained images were preprocessed using region of interest ( ROI ) segmentation, filtering denoising, and down-sampling. And then the depth and volume of processed images were obtained by the Oriented Bounding Box ( OBB ) algorithm and volume algorithms, separately. Two four-element viscoelastic models were established to fit depth and volume respectively for obtaining viscoelastic characteristics. Finally, regression models were built and compared using viscoelastic characteristics to determine the optimum prediction models and methods for two freshness indicators. The backpropagation neural network ( BPNN ) and SVR based on selected features were the best prediction models for pH and total volatile basic nitrogen ( TVB-N ) content evaluation in beef, and the correlation coefficients ( R c and R p ) of the calibration set and prediction set were 0.7636, 0.9036, and 0.7669, 0.8388, respectively. • A test electronic instrument based on airflow-3D machine vision was independently developed. • The depth and volume of spatial deformation were quantified by algorithms of point cloud image quantization. • Two four-element viscoelastic models of the creep recovery phase were established to extract viscoelastic characteristics. • Machine learning models for simultaneously predicting two common freshness indicators were built. • The prediction effects of different viscoelastic feature fusions and different models were compared.

Sustainable Utilization of Machine-Vision-Technique-Based Algorithm in Objective Evaluation of Confocal Microscope Images

Confocal microscope imaging has become popular in biotechnology labs. Confocal imaging technology utilizes fluorescence optics, where laser light is focused onto a specific spot at a defined depth in the sample. A considerable number of images are produced regularly during the process of research. These images require methods of unbiased quantification to have meaningful analyses. Increasing efforts to tie reimbursement to outcomes will likely increase the need for objective data in analyzing confocal microscope images in the coming years. Utilizing visual quantification methods to quantify confocal images with naked human eyes is an essential but often underreported outcome measure due to the time required for manual counting and estimation. The current method (visual quantification methods) of image quantification is time-consuming and cumbersome, and manual measurement is imprecise because of the natural differences among human eyes’ abilities. Subsequently, objective outcome evaluation can obviate the drawbacks of the current methods and facilitate recording for documenting function and research purposes. To achieve a fast and valuable objective estimation of fluorescence in each image, an algorithm was designed based on machine vision techniques to extract the targeted objects in images that resulted from confocal images and then estimate the covered area to produce a percentage value similar to the outcome of the current method and is predicted to contribute to sustainable biotechnology image analyses by reducing time and labor consumption. The results show strong evidence that t-designed objective algorithm evaluations can replace the current method of manual and visual quantification methods to the extent that the Intraclass Correlation Coefficient (ICC) is 0.9.

Machine Vision-Based Hook Defect Detection

Due to the harsh working environment of crane equipment, regular safety inspection is essential. The hook is one of the most frequently used parts of crane equipment. Therefore, if the hook lacks regular and standardized safety inspection, it can easily cause casualties. However, traditional machine vision techniques still face many challenges, making it difficult to extract the target object from the complex scene. In this paper, we propose a combination of machine vision techniques based on deep learning and traditional methods to extract hooks from complex construction site scenes and detect whether the hooks meet the criteria for proper operation.

Computer aided detection of mercury heavy metal intoxicated fish: an application of machine vision and artificial intelligence technique

Computer aided detection of mercury heavy metal intoxicated fish: an application of machine vision and artificial intelligence technique

An Objective Metallographic Analysis Approach Based on Advanced Image Processing Techniques

Metallographic analyses of nodular iron casting methods are based on visual comparisons according to measuring standards. Specifically, the microstructure is analyzed in a subjective manner by comparing the extracted image from the microscope to pre-defined image templates. The achieved classifications can be confused, due to the fact that the features extracted by a human being could be interpreted differently depending on many variables, such as the conditions of the observer. In particular, this kind of problem represents an uncertainty when classifying metallic properties, which can influence the integrity of castings that play critical roles in safety devices or structures. Although there are existing solutions working with extracted images and applying some computer vision techniques to manage the measurements of the microstructure, those results are not too accurate. In fact, they are not able to characterize all specific features of the image and, they cannot be adapted to several characterization methods depending on the specific regulation or customer. Hence, in order to solve this problem, we propose a framework to improve and automatize the evaluations by combining classical machine vision techniques for feature extraction and deep learning technologies, to objectively make classifications. To adapt to the real analysis environments, all included inputs in our models were gathered directly from the historical repository of metallurgy from the Azterlan Research Centre (labeled using expert knowledge from engineers). The proposed approach concludes that these techniques (a classification under a pipeline of deep neural networks and the quality classification using an ANN classifier) are viable to carry out the extraction and classification of metallographic features with great accuracy and time, and it is possible to deploy software with the models to work on real-time situations. Moreover, this method provides a direct way to classify the metallurgical quality of the molten metal, allowing us to determine the possible behaviors of the final produced parts.

A New Approach for Estimation of Physical Properties of Irregular Shape Fruit

Dimensions, volume, and mass of agricultural goods are essential physical features to build sizing, grading, and packaging systems. Manual property measurements are time-consuming, costly, and labor-intensive due to traditional technologies while various devices need to be installed to adapt process requirements. Furthermore, it is very challenging to accurately measure product with irregular or specific shapes, such as starfruit (Averrhoa carambola). Recently, there have been several research results on estimating features of the irregularly shaped object, they are either get inaccurate results or need repeated captures, computational resources, and time to rebuild the three-dimensional representation of the goods. The starfruit has not been studied completely in this size and weight measurements. This paper focuses on new techniques which exhibit simple installation to generate multiple functions for estimating the dimensions, volume, and mass with high accuracy. In this proposed method, we separated the process into two main phases. In the first phase, a camera is used to capture a top-view image of a starfruit, then image processing and machine vision techniques are applied to process the acquisition image before the process slices numerically the starfruit into several pieces along the longitudinal axis and estimates the physical attributes of each pieces using disc method and conical frustum method. Its volume is the summation of the volume of each partial slice. In the next phase, the density is used to estimate the mass of starfruit since the correlation coefficient (R-squared) between the volume and mass of starfruit is nearly linear with 0.9205. The validated results are highly competitive with accuracy of about 99% for the volume and mass in 300 testing samples.

Automated Evaluation of Surface Roughness using Machine Vision based Intelligent Systems

Machine vision systems play a vital role in entirely automating the evaluation of surface roughness due to the hitches in the conformist system. Machine vision systems significantly abridged the ideal time and human errors for evaluation of the surface roughness in a nondestructive way. In this work, face milling operations are performed on aluminum and a total of 60 diverse cutting experiments are conducted. Surface images of machined components are captured for the development of machine vision systems. Images captured are processed for texture features namely RGB (Red Green Blue), GLCM (Grey Level Co-occurrence Matrix) and an advanced wavelet known as curvelet transforms. Curvelet transforms are developed to study the curved textured lines present in the captured images and this module is capable to unite the discontinuous curved lines present in images. The CNC machined components consists of visible lay patterns in the curved form, so this novel machine vision technique is developed to identify the texture well over the other two extensively researched methods. Artificial Neural Network-Particle Swarm Optimization (ANN-PSO) intelligent models are developed to evaluate the surface roughness from texture features. The model average error attained using RGB, GLCM, Curvelet transform-based machine vision systems are 12.68, 7.8 and 3.57 respectively. In comparison, the results proved that computer vision system based on curvelet transforms outperformed the other two existing systems. This curvelet based machine vision system can be used for the evaluation of surface roughness. Here, image processing might be crucial in identifying certain information. One crucial issue is that, even as performance improves, cameras continue to get smaller and more affordable. The possibility for new applications in Industry 4.0 is made possible by this technological advancement and the promise of ever-expanding networking.

Equine Kinematic Gait Analysis Using Stereo Videography and Deep Learning: Stride Length and Stance Duration Estimation

Highlights Stereo machine vision and deep learning techniques were investigated for infield equine kinematic gait analysis. The proposed pipeline tracks equine body landmarks in 3D space and estimates stride length and stance duration. The system can serve as a cost-effective, rapid, and easy-to-use tool for equine locomotion research. Abstract. Equine kinematic gait analysis (EKGA) currently requires a complicated, expensive, and labor-intensive procedure for equine locomotion research. An automated stereo video processing pipeline was developed and evaluated for measuring equine biomechanical parameters. Using stereo videos of 40 different walking horses, a DeepLabCut (DLC) model was trained to detect body landmarks in individual frames. With an autoregressive integrated moving average filter, the landmark detection had a root mean square error of 5.14 pixels and a mean absolute error of 4.87 pixels. As a case study, methods were developed to extract stride length (SL) and stance duration (SD). Individual hoof gait phase detection was achieved using a fine-tuned Faster R-CNN model and a mode filter, yielding precision and recall values of 0.83 and 0.95, respectively. The semi-global block matching (SGBM) algorithm was used to estimate depth maps, and the accuracy was assessed by comparing head length estimation with infield measurements. A Bland-Altman analysis for DLC-detected head length in combination with SGBM-based 3D reconstruction yielded a bias of -0.014 m with upper and lower limits of agreement (LoAs) of 0.03 m and -0.061 m, respectively. Furthermore, Bland-Altman analyses on SD and SL when compared to image-level manual measurements showed biases of -0.02 sec and -0.042 m, respectively. The corresponding LoAs were (0.01907 sec, -0.24 sec) for SD and (0.04 m, -0.12 m) for SL. The proposed method showed promising potential in performing EKGA in an automated, cost-effective, and rapid manner under field conditions. Keywords: 3D Reconstruction, Animal Pose Estimation, Deep Learning, Equine Kinematic Gait Analysis, Stereo Matching.

Surface Characteristics Measurement Using Computer Vision: A Review

Computer vision provides image-based solutions to inspect and investigate the quality of the surface to be measured. For any components to execute their intended functions and operations, surface quality is considered equally significant to dimensional quality. Surface Roughness (Ra) is a widely recognized measure to evaluate and investigate the surface quality of machined parts. Various conventional methods and approaches to measure the surface roughness are not feasible and appropriate in industries claiming 100% inspection and examination because of the time and efforts involved in performing the measurement. However, Machine vision has emerged as the innovative approach to executing the surface roughness measurement. It can provide economic, automated, quick, and reliable solutions. This paper discusses the characterization of the surface texture of surfaces of traditional or non-traditional manufactured parts through a computer/machine vision approach and assessment of the surface characteristics, i.e., surface roughness, waviness, flatness, surface texture, etc., machine vision parameters. This paper will also discuss multiple machine vision techniques for different manufacturing processes to perform the surface characterization measurement.

Comparative and Evaluation of Anomaly Recognition by Employing Statistic Techniques on Humanoid Robot

This paper presents the study to differentiate between normal and anomaly conditions detected by humanoid robots using comparative statistics. The study has been conducted in robotic software as a platform to examine the scenario and evaluate between the anomalies and normal behaviour in different conditions. This study employed a machine vision technique to run an image segmentation process and carry out semi-supervised object training within a controlled environment. The robot is trained by differentiating the measurement size of the target object, its location, and the object’s visibility within three different frames. The effect is measured by extracting the positive predictive value (PPV) value, mean and standard deviation value from the captured image using statistical techniques in machine vision. The results showed that the mean value decreased by around 50% from the normal scenario when an anomaly occurred. Aside from that, the standard deviation values were more than twofold compared to the common scenario, especially after the object’s size grew. In contrast, the deviation value is remarkably small when the target is situated in the middle of adjacent frames, compared to the value when the entire shape is positioned in the frame. Simultaneously, the mean values from the processed image produced a minor difference.

Automatic extraction of data from infra-red thermometers with computer vision

Contactless infrared thermometers (NCITs) have been widely used as a rapid screening tool for patients with fever, particularly during the peak of the Covid-19 pandemic in Africa and other developing countries. The screening information that can be gathered in this manner is difficult to leverage from a pandemic response perspective as common handheld thermometers have no means of automatically exporting measurement data. An algorithm for a mobile phone-based application is developed that is capable of automatically extracting the temperature information from common NCITs by making use of machine vision techniques and partitioning tasks between the screening operator and the mobile application. Temperature readings could accurately be captured from a range of devices with varying text characteristics and backlights in a range of lighting conditions. Additionally the methodology proposed is capable of running on low-end devices without a data connection and without extensive training requirements for different NCITs being required. The use of a low-cost smart-phone as the capturing device together with the involvement of the operator has been show to be a feasible alternative to fully automating processes. Results indicate that similar methods could be used in a range of applications where numerical data needs to be digitised from existing field instrumentation, particularly where such instrumentation feature segmented displays.

A Review on Potato Disease Detection and Classification by exploiting Deep Learning Techniques

The edible potato comes in fifth for human consumption and fourth among main food crops. Since it is a crop that is vegetative grown, many pests and disease can be passed along from one generation to the next. Crop diseases, which have a negative impact on food security as well as economic losses, have a significant impact on the production and quality of yields from potato crops. Thus, the application of unique and precise deep learning-based algorithms for disease detection and classification is highly required. Identifying weaknesses in agricultural products, particularly potatoes requires the use of machine vision and image processing techniques. Deep learning and image processing have been used in agriculture to classify and number of disease and pests affecting potatoes has grown, and study in this area is still continuing. The use of artificial intelligence and image processing in agriculture for the classification and identification of potato pests and disease has grown, and work in this area is still ongoing. Different deep learning techniques, such as VGG19, VGG16, Google Net, Alex Net, and convolution neural network methods, can be used to address the disease problem in potatoes. These methods also examined multiple classes of potato diseases as: Healthy, Black Leg, Black Scurf, Pink Rot, Common Scab, etc. Food safety could be seriously threatened by the spread of potato disease. In this article, deep learning techniques for early detection of potato disease are discussed.

A Comparative Analysis of Machine Vision Algorithms for Insect Pests Classification

Insect pests are the one of the important biological factors, that has become an important cause of crop yield degradation. However, their identification and detection in the early stages is a very significant task to minimize the overall losses. The conventional techniques with naked eyes to identify the pests is very exigent and require domain specific expertise. It is extremely time-consuming and tedious task to identify the pests in the initial stages with conventional methods. To minimize these issues, some highly developed methods are required to detect insect pests accurately in agriculture. The continuous emergence of machine vision in image processing helps in this regard. This paper presents a comprehensive review to identify the insect pests in the early stages with the help of machine vision techniques. Based on this, a comparative analysis of different classifiers has also been presented.

Smart Agriculture: A Fruit Flower Cluster Detection Strategy in Apple Orchards Using Machine Vision and Learning

This paper presents the application of machine vision and learning techniques to detect and identify the number of flower clusters on apple trees leading to the ability to predict the potential yield of apples. A new field robot was designed and built to collect and build a dataset of 1500 images of apples trees. The trained model produced a cluster precision of 0.88 or 88% and a percentage error of 14% over 106 trees running the mobile vehicle on both sides of the trees. The detection model was predicting less than the actual amount but the fruit flower count is still significant in that it can give the researcher information on the estimated growth and production of each tree with respect to the actions applied to each fruit tree. A bias could be included to compensate for the average undercount. The resulting F1-Score of the object detection model was 80%, which is similar to other research methods ranging from an F1-Score of 77.3% to 84.1%. This paper helps lay the foundation for future application of machine vision and learning techniques within apple orchards or other fruit tree settings.

A Robust Faster R-CNN Model with Feature Enhancement for Rust Detection of Transmission Line Fitting.

Rust of transmission line fittings is a major hidden risk to transmission safety. Since the fittings located at high altitude are inconvenient to detect and maintain, machine vision techniques have been introduced to realize the intelligent rust detection with the help of unmanned aerial vehicles (UAV). Due to the small size of fittings and disturbance of complex environmental background, however, there are often cases of missing detection and false detection. To improve the detection reliability and robustness, this paper proposes a new robust Faster R-CNN model with feature enhancement mechanism for the rust detection of transmission line fitting. Different from current methods that improve feature representation in front end, this paper adopts an idea of back-end feature enhancement. First, the residual network ResNet-101 is introduced as the backbone network to extract rich discriminative information from the UAV images. Second, a new feature enhancement mechanism is added after the region of interest (ROI) pooling layer. Through calculating the similarity between each region proposal and the others, the feature weights of the region proposals containing target object can be enhanced via the overlaying of the object’s representation. The weight of the disturbance terms can then be relatively reduced. Empirical evaluation is conducted on some real-world UAV monitoring images. The comparative results demonstrate the effectiveness of the proposed model in terms of detection precision and recall rate, with the average precision of rust detection 97.07%, indicating that the proposed method can provide an reliable and robust solution for the rust detection.