The development of drone technology is very rapid. From the use of U.S. military drones to the popularization of UAVs in research and civilian use, UAVs have become a new trend [1-2]. Along with it, many new problems have also been brought about. The accident of the unmanned aircraft crashing into the fighter plane has given people a wake-up call. Therefore, the construction of the drone police system is imperative. The research focus of this paper is: to establish a visual sensor network for image capture and information storage of drones; to introduce deep learning to identify unmanned aerial vehicles, to find "black fly drones" in time, and to take corresponding alarm measures. Achieve comprehensive supervision of drones.
1, visual sensor network
The entire Visual Sensor Networks (VSNs) consists of multiple nodes, each of which will be composed of a camera array, which will serve as the basis for the entire system [3], as shown in Figure 1.
A schematic diagram of the placement of a node in an urban environment is shown in Figure 2.
In order to reduce the interference to the residents, the camera focal length parameter can be modified to limit the shooting range of the camera. Through the cross coverage of multiple cameras, the middle open space area was successfully integrated into the monitoring.
Considering the huge amount of data provided by multiple nodes and the need to optimize the control structure, the data network is designed into a three-layer structure. The third layer at the bottom layer consists of a series of clusters of nodes of varying numbers, and the nodes in each cluster uniformly send data to a secondary processing server. The secondary processing servers throughout the network form the second layer of the network, sending data to the central advanced server at the first tier.
2. Image recognition center based on deep learning
The key component of the drone police system is the image recognition center. Its task is to analyze and process the image information in the visual sensor network, identify the drone from the image, and realize the monitoring of the drone. Identify the field. There have been a lot of outstanding achievements in this field. The most common pedestrian detection problems, available features include: Haar, HOG, CSS, LBP, etc. These features express important parts of the human body, and fully consider the occlusion and other situations. Wang Xiaogang and Ouyang Wanli have proposed a pedestrian detection method based on deep learning, which maximizes their respective roles by jointly learning the four important components of pedestrian detection—feature extraction, body part deformation processing, occlusion processing and classification [4]. ]. On the basis of the traditional convolutional neural network, they added the deformation processing layer, and the finally acquired features have strong discriminative power, which is superior to HOG and other features. The program of Wang Xiaogang's team is the successful application of deep learning in the field of target recognition, which provides a research reference for the research of this paper. Another example is the face recognition problem [5-6], which has more complicated changes, because the face is affected by many factors such as race, skin color, expression, emotion, lighting environment, and object occlusion. There are also many schemes for promotion to the recognition of various specific objects, as well as scene recognition and deep learning [7]. Due to the rich amount of picture information in the drone police system and the variety of flight status of the drone, it is difficult to identify. To this end, this paper will introduce a deep learning algorithm and use convolutional neural networks as the image recognition center.
2.1 Convolutional neural networks
In 2006, Hinton et al. first proposed the concept of deep learning [8], and opened up a wave of deep learning research. It believes that multi-hidden artificial neural networks can better simulate the thinking process of the human brain, and have more excellent Learning ability, the ability to more fundamentally characterize the data, thereby improving the ability to visualize or classify.
Convolutional neural network is the first real multi-layer structure learning algorithm in deep learning, which has obvious advantages in the field of image recognition. It uses the concepts of receptive field and local connection to greatly reduce the parameter quantity, reduce the complexity of the network model, improve the training efficiency, and the network is highly invariant to various deformations of translation and zoom.
Convolutional neural networks are a kind of feedforward multi-layer neural network. Each layer is composed of multiple two-dimensional planes. Multiple neurons form each plane. The structure is shown in Figure 3.
The convolutional neural network utilizes a series of convolutional layers. The downsampling layer constructs a multi-layer network to simulate the layer-by-layer processing mechanism of human brain perception visual signals, thereby extracting multi-level features of the image.
By adding a convolution layer, a local connection network can be realized, which effectively reduces the network parameters that need to be trained. For example, for a large picture input, the size is r&TImes; c, the random sample is a&TImes; b small picture, if the hidden node is k, then the number of features finally learned is:
The convolutional neural network utilizes a series of convolutional layers. The downsampling layer constructs a multi-layer network to simulate the layer-by-layer processing mechanism of human brain perception visual signals, thereby extracting multi-level features of the image.
VEIIK Vape is so convenient, portable, and small volume, you just need to take them out of your pocket and take a puff, feel the cloud
of smoke, and the fragrance of fruit surrounding you. It's so great.
We are the distributor of the VEIIK Vape brand, we sell veiik e cigarette, veiik vape pen, veiik disposable vaporizer, and so on.
We are also China's leading manufacturer and supplier of Disposable Vapes puff bars, disposable vape kit, e-cigarette, vape pens, and e-cigarette kit,
and we specialize in disposable vapes, e-cigarette vape pens, e-cigarette kits, etc.
veiik vape disposable,veiik vape pen,veiik e cigarette,veiik vape kit,veiik disposable vaporizer
Ningbo Autrends International Trade Co.,Ltd. , https://www.supermosvape.com