In this paper, we propose a data-driven proposal and deep-learning based classification scheme for small targetdetection. Previous studies have shown feasible performance using conventional computer vision techniques...
详细信息
ISBN:
(纸本)9781510617605
In this paper, we propose a data-driven proposal and deep-learning based classification scheme for small targetdetection. Previous studies have shown feasible performance using conventional computer vision techniques such as spatial and temporal filters. However, those are handcrafted approaches and are not optimized due to the nature of the application fields. Recently, deep learning has shown excellent performance for many computer vision problems, which motivates the deep learning-based small targetdetection. The proposed data-driven proposal and convolutional neural network (DDP-CNN) can generate possible target locations through the data-driven proposal and final targets are recognized through the classification network. According to the experimental results using drone database, the DDP-CNN shows 91% of train accuracy and 0.85 of average precision (AP) of the targetdetection.
Hyperspectral images contain hundreds of continuous spectral bands, which can provide rich information for targetdetection and image classification. However, high-dimensional data are easy to cause the "Hughes p...
详细信息
Hyperspectral images contain hundreds of continuous spectral bands, which can provide rich information for targetdetection and image classification. However, high-dimensional data are easy to cause the "Hughes phenomenon", which affects the classification accuracy of images. In this paper, a band selection method based on multi-criteria ranking is proposed for hyperspectral dimensionality reduction. First, the hyperspectral image acquisition system is built to collect the hyperspectral transmission image datasets containing various heterogeneities. Second, the original data are preprocessed and the band selection method proposed is used for data dimensionality reduction. That is, (1) the whole band is divided into subintervals according to the correlation criteria, and then the bands with large information amount and good inter-class separability are selected according to other criteria functions, (2) the bands of each subinterval are combined, and the combined bands are evaluated according to the optimum index factor and Bhattacharyya distance, and (3) the combined bands are ranked and the top-ranked combined bands are selected as the representative bands for image classification. Finally, the support vector machine is used to verify the effectiveness of the representative bands selected by this method. The experimental results show that the proposed method achieves data dimensionality reduction, and the overall classification accuracy can reach 100% on some selected subsets of bands, which has better classification performance than other related band selection methods. In addition, the computational complexity of high-dimensional data can be effectively reduced by band selection, and the average running time is 95.19% less than that of the full bands while ensuring classification accuracy.
The infrared thermal imager is a common method for electric inspection to detect the abnormal heat of the gold tool at the tensioning wire clamp. Because the Gansu section of Qishao High voltage line accounts for abou...
详细信息
The infrared thermal imager is a common method for electric inspection to detect the abnormal heat of the gold tool at the tensioning wire clamp. Because the Gansu section of Qishao High voltage line accounts for about 52.4% of the total length, through a lot of complex geographical environment, even once a quarter manual inspection will cost time and effort, subjectivity is big, and real-time performance is poor. Aiming at the above, this paper proposes an improved YOLOv4 multi-target recognition algorithm for infrared images. Firstly, the backbone feature extraction network CSPDaeknet53 is replaced by Mobilenetv1. Secondly, the training dataset is increased by the base hjs image expansion technology. Finally, Mosaic data enhancement and cosine annealing attenuation technology are introduced to construct an effective infrared image detection model to realize the classification and recognition of the hardware. The experimental results are compared with SSD, YOLOv3 and YOLOv4 targetdetection algorithms, and it is found that the average recognition accuracy of the proposed method is 93.84%, and the model size is only 55.53MB. Under the condition of ensuring accuracy, the model is smaller and has practical reference value.
infrared image recognition in substation is always a difficult problem. In order to solve the problem of recognition of knife gate, insulator and other components in infrared image, the target location technology of d...
详细信息
ISBN:
(数字)9781728176871
ISBN:
(纸本)9781728176871
infrared image recognition in substation is always a difficult problem. In order to solve the problem of recognition of knife gate, insulator and other components in infrared image, the target location technology of deep learning is proposed to realize the detection and recognition of typical components in infrared image, and the multi-targetdetection algorithm YOLO is selected to locate and identify the defects. Firstly, the image preprocessing technology is used to process the collected image, so as to filter the interference of background and other factors on the equipment identification. Then, the infrared image is detected by the YOLO targetdetection model based on multi feature fusion, so as to locate the position of inspection equipment in the infrared image. Then, the type of equipment is identified by the trained equipment classification model. Finally, the algorithm is tested with a large number of pictures in the substation scene.
Honey is a highly desirable commodity hence a target of adulteration to increase its bulk. Spectroscopic methods and chemometrics offer a fast, easy, and simple approach to the detection of syrup adulterations in hone...
详细信息
Honey is a highly desirable commodity hence a target of adulteration to increase its bulk. Spectroscopic methods and chemometrics offer a fast, easy, and simple approach to the detection of syrup adulterations in honey. The idiosyncrasies associated with different instrumental data suggest varying performance for each algorithm. The objective of this study was to evaluate the performance of five commonly used classification and regression algorithms for the detection of syrup adulteration in honey. The prediction performance of the classification and regression algorithms was evaluated using data obtained from Fourier-transform infrared spectrometer with a Horizontal Attenuated Total Reflectance (HATR) Accessory analysis of honey samples. Gradient boosted discriminant analysis (GBDA) and Support vector machines discriminant analysis (SVMDA) were able to differentiate between adulterated and pure honey samples with an external validation set prediction accuracies of 0.988 and 0.981, respectively. Application of feature selection method did not lead to improved prediction accuracies. Gradient boosted regression (GBR) initialized with a ridge regression predicted the percentage level of adulteration with a regression coefficient of 1.000. The RMSE for the optimal GBR was 2.183 and 0.018 for before and after feature selection with the partial least squares (PLS) algorithm, respectively. Ensemble methods are generally better for both classification and regression of honey using FTIR-HATR spectroscopic data.
暂无评论