This study explores traditional object detection algorithms and deep neural network-based engineering vehicle detection algorithms. We applied preprocessing algorithms such as image denoising, enhancement, and edge de...
详细信息
ISBN:
(纸本)9798400709272
This study explores traditional object detection algorithms and deep neural network-based engineering vehicle detection algorithms. We applied preprocessing algorithms such as image denoising, enhancement, and edge detection to process the images and constructed our own trainable dataset through labeling. Next, we used the YOLOv5 algorithm, which has high detection accuracy and real-time capability, for engineering vehicle detection. To address the issues of missed targets and predicted bounding box misalignment, we introduced the DeepSORT algorithm for target prediction and tracking. This algorithm utilizes Kalman filtering for estimation and updates and employs the Hungarian algorithm to associate data between consecutive frames, thereby achieving engineering vehicle tracking. To tackle the problem of frequent identity switching due to camera motion and non-uniform vehicle movements, we adopted the modified GIoU to calculate the intersection over union between trajectories and detected target bounding boxes, reducing identity jumps during the tracking process. Finally, we designed an engineering vehicle detection and tracking system and applied this algorithm to practical production.
A new fault diagnosis method based on ITD sample entropy and probabilistic neural network is proposed. Firstly, the vibration signal of the equipment is decomposed using intrinsic time-scale decomposition (ITD) to obt...
详细信息
Accurate prediction of aero-engine remaining useful life (RUL) is essential for providing reliable maintenance or alarm decisions. The extraction of degraded features has a great impact on the accuracy of RUL predicti...
详细信息
With the development of hyperspectral sensors, there is an increasing amount of accessible hyperspectral data, and the classification task for land cover categories has gained significant attention. Existing classific...
With the development of hyperspectral sensors, there is an increasing amount of accessible hyperspectral data, and the classification task for land cover categories has gained significant attention. Existing classification methods typically extract features from either the pixel or superpixel perspective. However, using a single-scale feature extraction approach fails to simultaneously consider both local and global features of land cover, leading to suboptimal classification results. To address this issue, this paper proposes a parallel graph attention network model based on pixel and superpixel feature fusion (SSPGAT) for hyperspectral image classification, which leverages the fusion of pixel-level and superpixel-level features. The proposed approach first employs spectral convolutional layers to reduce the redundant spectral dimension. Then, it utilizes graph attention network (GAT) to extract local and global features of land cover separately from the pixel and superpixel perspectives. Finally, a fully connected network is employed to classify the fused features from both branches. Experimental results on two different datasets demonstrate the effectiveness of the proposed approach.
A reconfigurable feed network for multi-polar antennas is proposed in this article. It uses PIN diodes to make reconfigurable structure and change the output. The network can produce four kinds of output signal with o...
详细信息
This paper studies the visual navigation of vehicles in unknown environments. Due to the lack of global information, the accuracy of graphical navigation method will be decline. We propose a double LSTM attention navi...
详细信息
Currently, a large number of community integrated energy systems operate independently and lack coordination and cooperation with each other. On the whole, the complementary coupling potential between energy productio...
详细信息
The existing wrapper feature selection methods have problems such as the difficulty of balancing a relatively high classification accuracy rate with a low feature selection rate, and there are problems such as long ti...
详细信息
ISBN:
(数字)9798331540043
ISBN:
(纸本)9798331540050
The existing wrapper feature selection methods have problems such as the difficulty of balancing a relatively high classification accuracy rate with a low feature selection rate, and there are problems such as long time consumption. In response to these issues, this paper combines quantum computing theory and sand cat swarm optimization to design quantum sand cat swarm optimization and combined with K-nearest neighbor method for feature selection. The optimal solution of quantum sand cat swarm optimization is set as the selected optimal feature subset. Simulation experiments are executed on 6 UCI data sets to compare with 4 classical swarm intelligence algorithms. The simulation outcomes show that the wrapper feature selection on the ground of quantum sand cat swarm optimization achieves lower feature selection rate , faster time and higher classification accuracy than the comparison method. It is a high precision and practical wrapper feature selection method.
Airborne radars usually face non-uniform clutter environments, and it is difficult to obtain enough independent and identical distributed (i.i.d.) training samples, which degrades the clutter suppression performance o...
详细信息
The Radio Frequency Fingerprint Identification (RFFI) technique utilizes the subtle and unintentional modulation present in the transmitted RF waveform, caused by the non-ideal characteristics of the device, to unique...
详细信息
暂无评论