First, photoelectric and radar cooperation in target recognition mainly has the following ways:
1. Data fusion and complementary identification
Data layer fusion: the target distance, speed, azimuth and other data obtained by radar are fused with the target image, color, texture and other data obtained by the photoelectric system at the data level. For example, by using Kalman filter and other algorithms, the two kinds of data are weighted and fused to obtain more accurate target state information, which provides a more comprehensive data basis for subsequent identification.
Feature layer fusion: Extract features from radar data and photoelectric image data, such as radar scattering cross-sectional area features, Doppler frequency features, and photoelectric shape, edge, temperature distribution and other features, and then fuse these features to form a more representative and distinguishable joint feature vector for target classification and recognition.
Decision level fusion: the radar and photoelectric system independently identify and judge the target, and then merge the recognition results of the two to make a decision. For example, voting method, Bayesian reasoning and other methods are used to comprehensively consider the recognition results of the two sensors to improve the accuracy and reliability of target recognition.
2. Target positioning and guidance recognition
Radar guidance photoelectric: the radar first carries out long-distance detection and rough positioning of the target, determines the approximate orientation and distance of the target, and then guides the optical lens or infrared detector of the photoelectric system to point at the target, so that the photoelectric system can quickly and accurately capture the image information of the target, and achieve accurate positioning and detailed identification of the target.
Photoelectric assisted radar: In some complex scenes, when the radar is disturbed or the target signal is weak, the photoelectric system can provide the initial position or auxiliary positioning information of the target for the radar through the imaging recognition of the target, help the radar search and track the target more effectively, and improve the radar's target acquisition capability.
Classification and recognition of complementary advantages
Use the radar's all-weather capability: the radar is not limited by light, weather and other conditions, and can detect and identify targets in harsh environments. For example, in the night, fog, rainy and other conditions, the radar can work normally, through the electromagnetic reflection characteristics of the target to classify it, such as to distinguish different types of aircraft, vehicles and so on.
The photoelectric system has a high-resolution imaging capability and can obtain detailed appearance characteristics and texture information of the target. For some targets with obvious visual characteristics, such as specific types of aircraft, ships, etc., the photoelectric image can be accurately classified and identified. At the same time, the photoelectric system can also use multi-spectral imaging technology to identify the camouflage target.
3. Dynamic monitoring and collaborative identification
Relay tracking and identification: During the movement of the target, the radar and photoelectric system can carry out relay tracking. When the target is far away, the radar continues to track the target, and as the target gradually approaches, the photoelectric system relays for more accurate tracking and recognition, giving full play to the advantages of the two in different distance segments to achieve stable recognition of the target throughout the process.
Joint state estimation: Through the joint processing of data acquired by radar and photoelectric system, real-time estimation of the target's motion state and attitude change. For example, the speed information of the radar and the image sequence of the photoelectric system are used to analyze the movement trajectory and turning Angle of the target, so as to identify the behavior pattern and intention of the target more accurately, and improve the dynamic recognition ability of the target.
4. Joint recognition based on machine learning
Joint training model: Radar and photoelectric data are used as inputs to jointly train machine learning models, such as neural networks and support vector machines, so that the model learns the intrinsic association and feature combination between the two data, so as to achieve more accurate target recognition. By training and optimizing the model with a large amount of labeled data, the recognition ability and generalization ability of the model to different types of targets can be improved.
Transfer learning application: Transfer the knowledge and feature representation of a model trained on one sensor data to another sensor data to assist target recognition.
Practical application of cooperative mode of radar in target recognition
1. Field of military air defense
Data fusion and complementary identification applications: In air defense systems, radars can detect incoming air targets, such as fighter jets and cruise missiles, at long distances. The radar can quickly acquire information such as distance (such as 300 km away), speed (such as Mach 2 flight speed) and azimuth (such as azimuth is 30 degrees north by east) of the target. At the same time, photoelectric systems (such as infrared thermal cameras and optical cameras) point in the direction of the target according to the data provided by the radar. The photoelectric system obtains high-resolution images of the target, for example, the thermal characteristics of the target engine can be found through the infrared thermal imager, and the details such as the outline and wing shape of the target can be seen through the optical camera. The data of radar and photoelectric system are fused in the feature layer, and the joint feature vector is constructed by the electromagnetic and optical characteristics of the target, which is used to distinguish the attacking target from the enemy fighter or the decoy bomb.
Target location and guidance identification example: When the early warning radar finds a suspicious target approaching the airspace, it will guide the photoelectric system of the air defense position to lock the target. For example, in the air defense facilities in the border area, the radar first detects the target's activity in the airspace 200 kilometers away from the border line, and sends the target's orientation information to the photoelectric system. The optical and infrared devices of the photoelectric system quickly adjust the pointing and image the target. The photoelectric system can identify that the target is a fighter with a specific shape, and then combine the distance and speed information of the radar to judge whether it is threatening, so as to provide a basis for air defense decisions.
Dynamic monitoring and cooperative identification scenario: During the flight of the target, the radar continuously tracks the position and speed changes of the target. When the target enters the effective identification range of the photoelectric system (such as within 50 kilometers from the defensive position), the photoelectric system relays for more detailed identification. For example, an electro-optical system can further determine a target's attack intent by looking at the type of weapon mounted under the target's wing. At the same time, through the joint processing of radar and photoelectric system data, analyze the maneuvering actions of the target, such as diving, circling, etc., real-time estimate the threat degree of the target, and provide dynamic information for the interception decision of the air defense system.
2. Maritime ship monitoring field
Complementary classification identification case: In the monitoring of ships in ports or sea areas, radar can work in all-weather conditions. On foggy days, the radar can detect ship targets in the surrounding waters and distinguish large cargo ships from small fishing boats by analyzing electromagnetic characteristics such as the radar scattering cross-sectional area of the targets. When the weather turns nice, photoelectric systems such as Marine optical telescopes and infrared thermal imagers come into play. The photoelectric system can obtain the detailed appearance information of the ship, such as the type of ship, the name of the ship, the shape of the superstructure, etc., and can also observe the power system of the ship through infrared imaging. Using the radar's all-weather detection capability and the photoelectric system's high-resolution imaging advantages, ships can be accurately classified and identified, such as identifying oil tankers, container ships or warships.
Joint recognition practice based on machine learning: In a maritime traffic management system, data from radar and electro-optical systems are used as inputs to train machine learning models. A large number of ship radar echo data and corresponding photoelectric image data are collected to train the model. For example, through the support vector machine model, the association between the electromagnetic characteristics of the ship in the radar data and the appearance characteristics in the photoelectric image is learned. When a new target appears, the model can accurately identify the type, size and sailing status of the ship according to the data provided by the radar and photoelectric system, improving the efficiency and accuracy of maritime ship monitoring.
3. Intelligent transportation
Data fusion and complementary recognition Example: In the environmental awareness system of an autonomous vehicle, the on-board radar can detect the distance of the surrounding vehicles (such as the vehicle in front of the car 10 meters away from the car), speed (such as the relative speed of 5 meters/second) and other information. At the same time, the on-board photoelectric system (such as the camera) can obtain the appearance image of the vehicle, including the license plate number, vehicle color, model and so on. Through data fusion, the radar range velocity feature and photoelectric appearance feature are combined in the feature layer to construct the joint feature vector. By using this joint feature vector, the surrounding vehicles can be more accurately identified as ordinary cars, SUVs or trucks, and the relative position relationship between vehicles can be judged in real time, providing more accurate environmental information for the decision-making system of autonomous vehicles.
Relay tracking and identification application: In the intelligent traffic monitoring system, for the vehicle monitoring on the highway, the long-range radar can track the vehicle and obtain the vehicle's driving trajectory and speed information. Through the relay tracking and identification of radar and photoelectric system, the effective monitoring of traffic flow and accurate management of vehicles are realized.