Search results

The purpose of this paper is to monitor the backup indicators in case of indicator failure and to minimize the situations when the pilot may be unable to monitor the indicator effectively in emergency situations.

In this study, the pointer positions of different indicators were determined with a deep learning-based algorithm. Within the scope of the study, the pointer on the analog indicators obtained from aircraft cockpits was detected with the YOLOv4 object detector. Then, segmentation was made with the GrabCut algorithm to detect the pointer in the detected region more precisely. Finally, a line including the segmented pointer was found using the least-squares method, and the exact direction of the pointer was determined and the angle value of the pointer was obtained by using the inverse tangent function. In addition, to detect the pointer of the YOLOv4 object detection method and to test the designed method, a data set consisting of videos taken from aircraft cockpits was created and labeled.

The analog indicator pointers were detected with great accuracy by the YOLOv4 and YOLOv4-Tiny detectors. The experimental results show that the proposed method estimated the angle of the pointer with a high degree of accuracy. The developed method can reduce the workloads of both pilots and flight engineers. Similarly, the performance of pilots can be evaluated with this method.

The authors propose a novel real-time method which consists of detection, segmentation and line regression modules for mapping the angle of the pointers on analog indicators. A data set that includes analog indicators taken from aircraft cockpits was collected and labeled to train and test the proposed method.

The advancements of deep learning (DL) models demonstrate significant performance on accurate pancreatic tumor segmentation and classification.

The presented model involves different stages of operations, namely preprocessing, image segmentation, feature extraction and image classification. Primarily, bilateral filtering (BF) technique is applied for image preprocessing to eradicate the noise present in the CT pancreatic image. Besides, noninteractive GrabCut (NIGC) algorithm is applied for the image segmentation process. Subsequently, residual network 152 (ResNet152) model is utilized as a feature extractor to originate a valuable set of feature vectors. At last, the red deer optimization algorithm (RDA) tuned backpropagation neural network (BPNN), called RDA-BPNN model, is employed as a classification model to determine the existence of pancreatic tumor.

The experimental results are validated in terms of different performance measures and a detailed comparative results analysis ensured the betterment of the RDA-BPNN model with the sensitivity of 98.54%, specificity of 98.46%, accuracy of 98.51% and F-score of 98.23%.

The study also identifies several novel automated deep learning based approaches used by researchers to assess the performance of the RDA-BPNN model on benchmark dataset and analyze the results in terms of several measures.

Extraction of leaf region from the plant leaf images is a prerequisite process for species recognition, disease detection and classification and so on, which are required for crop management. Several approaches were developed to implement the process of leaf region segmentation from the background. However, most of the methods were applied to the images taken under laboratory setups or plain background, but the application of leaf segmentation methods is vital to be used on real-time cultivation field images that contain complex backgrounds. So far, the efficient method that automatically segments leaf region from the complex background exclusively for black gram plant leaf images has not been developed.

Extracting leaf regions from the complex background is cumbersome, and the proposed PLRSNet (Plant Leaf Region Segmentation Net) is one of the solutions to this problem. In this paper, a customized deep network is designed and applied to extract leaf regions from the images taken from cultivation fields.

The proposed PLRSNet compared with the state-of-the-art methods and the experimental results evident that proposed PLRSNet yields 96.9% of Similarity Index/Dice, 94.2% of Jaccard/IoU, 98.55% of Correct Detection Ratio, Total Segmentation Error of 0.059 and Average Surface Distance of 3.037, representing a significant improvement over existing methods particularly taking into account of cultivation field images.

In this work, a customized deep learning network is designed for segmenting plant leaf region under complex background and named it as a PLRSNet.

The purpose of this paper is to present an efficient, interactive foreground/background image segmentation method using mean shift (MS) and graph cuts, in order to improve the segmentation performance with little user interaction.

By incorporating the advantages of the mean shift method and the graph cut algorithm, the proposed approach ensures the accuracy of segmentation results. First, the user marks certain pixels as foreground or background. Then the graph is constructed and the cost function composed of the boundary properties and the region properties is defined. To obtain the hidden information of user interaction, the foreground and background marks are clustered separately by the mean shift method. The region properties are determined by the minimum distances from the unmarked pixels to the foreground and background clusters. The boundary properties are determined by the relationship between the unmarked pixels and its neighbor pixels. Finally, using the graph cuts method solves the energy minimization problem to get the foreground which is of interest.

The paper presents experimental results and compares the results to other methods. It can be seen from the comparison that this method can obtain a better segmentation performance in many cases.

The paper incorporates the advantages of the mean shift method and the graph cut algorithm to obtain better segmentation results, even though the scene is complex.

The purpose of this paper is to explore and examine discrepancies of food aesthetics portrayed on social media across different types of restaurants using a large-scale data set of food images.

A neural food aesthetic assessment model using computer vision and deep learning techniques is proposed, applied and evaluated on the food images data set. In addition, a set of photographic attributes drawn from food services and cognitive science research, including color, composition and figure–ground relationship attributes is implemented and compared with aesthetic scores for each food image.

This study finds that restaurants with different rating levels, cuisine types and chain status have different aesthetic scores. Moreover, the authors study the difference in the aesthetic scores between two groups of image posters: customers and restaurant owners, showing that the latter group tends to post more aesthetically appealing food images about the restaurant on social media than the former.

Restaurant owners may consider performing more proactive social media marketing strategies by posting high-quality food images. Likewise, social media platforms should incentivize their users to share high-quality food images.

The main contribution of this paper is to provide a novel methodological framework to assess the aesthetics of food images. Instead of relying on a multitude of standard attributes stemming from food photography, this method yields a unique one-take-all score, which is more straightforward to understand and more accessible to correlate with other target variables.

The purpose of this paper is to propose a new video prediction-based methodology to solve the manufactural occlusion problem, which causes the loss of input images and uncertain controller parameters for the robot visual servo control.

This paper has put forward a method that can simultaneously generate images and controller parameter increments. Then, this paper also introduced target segmentation and designed a new comprehensive loss. Finally, this paper combines offline training to generate images and online training to generate controller parameter increments.

The data set experiments to prove that this method is better than the other four methods, and it can better restore the occluded situation of the human body in six manufactural scenarios. The simulation experiment proves that it can simultaneously generate image and controller parameter variations to improve the position accuracy of tracking under occlusions in manufacture.

The proposed method can effectively solve the occlusion problem in visual servo control.

The purpose of this paper is to develop an automated human body measurement extraction system using simple inexpensive equipment with minimum requirement of human assistance. This research further leads to the comparison of extracted measurements to established methods to analyze the error. The extracted measurements can be used to assist the production of custom-fit apparel. This is an effort to reduce the cost of expensive 3-D body scanners and to make the system available to the user at home.

A single camera body measurement system is proposed, implemented, and pilot tested. This system involves a personal computer and a webcam operating within a space of controlled lighting. The system will take two images of the user, extract body silhouettes, and perform measurement extraction. The camera is automatically calibrated using the software each time of scanning considering the scanning space. The user will select a front view and a side view among the images captured, and specify the height. In this pilot study, 31 subjects were recruited and the accuracy of 8 human body measurements were compared with the manual measurements and measurements extracted from a commercial 3-D body scanner.

The system achieved reasonable measurement performance within 10 percent accuracy for seven out of the eight measurements, while four out of eight parameters obtained a performance similar to the commercial scanner. It is proved that human body measurement extraction can be done using inexpensive equipment to obtain reasonable results.

This study is aimed at developing a proof-of-concept for inexpensive body scanning system, with an effort to benchmark measurement accuracy, available to an average user providing the ability to acquire self-body measurements to be used to purchase custom-fit apparel. This system can potentially boost the customization of apparel and revolutionize online shopping of custom-fit apparel.

Laser-based measurement techniques offer various advantages over conventional measurement techniques, such as no-destructive, no-contact, fast and long measuring distance. In cooperative laser ranging systems, it’s crucial to extract center coordinates of retroreflectors to accomplish automatic measurement. To solve this problem, this paper aims to propose a novel method.

We propose a method using Mask RCNN (Region Convolutional Neural Network), with ResNet101 (Residual Network 101) and FPN (Feature Pyramid Network) as the backbone, to localize retroreflectors, realizing automatic recognition in different backgrounds. Compared with two other deep learning algorithms, experiments show that the recognition rate of Mask RCNN is better especially for small-scale targets. Based on this, an ellipse detection algorithm is introduced to obtain the ellipses of retroreflectors from recognized target areas. The center coordinates of retroreflectors in the camera coordinate system are obtained by using a mathematics method.

To verify the accuracy of this method, an experiment was carried out: the distance between two retroreflectors with a known distance of 1,000.109 mm was measured, with 2.596 mm root-mean-squar error, meeting the requirements of the coarse location of retroreflectors.

The research limitations/implications are as follows: (i) As the data set only has 200 pictures, although we have used some data augmentation methods such as rotating, mirroring and cropping, there is still room for improvement in the generalization ability of detection. (ii) The ellipse detection algorithm needs to work in relatively dark conditions, as the retroreflector is made of stainless steel, which easily reflects light.

The originality/value of the article lies in being able to obtain center coordinates of multiple retroreflectors automatically even in a cluttered background; being able to recognize retroreflectors with different sizes, especially for small targets; meeting the recognition requirement of multiple targets in a large field of view and obtaining 3 D centers of targets by monocular model-based vision.