Intelligent monitoring system and control method thereof
By using multimodal feature fusion and spatiotemporal graph convolutional networks, the problems of reduced recognition performance and target association in intelligent monitoring systems under complex environments are solved, achieving highly stable target detection and accurate group anomaly recognition, thus improving the proactive early warning capability of security systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEILONGJIANG NORTH TOOLS CO LTD
- Filing Date
- 2025-11-28
- Publication Date
- 2026-04-17
AI Technical Summary
Existing intelligent monitoring systems suffer from a sharp decline in recognition performance in complex environments such as low light, rain, and fog. The recall rate for pedestrian detection at night is less than 60%. Traditional detection algorithms struggle to understand the spatiotemporal relationships between targets, and the accuracy rate for identifying abnormal group behavior is less than 75%. Centralized cloud computing architectures also pose risks of high network latency and data privacy leaks.
A distributed intelligent monitoring architecture based on multimodal feature fusion is constructed. It adopts multi-source data complementarity from visible light, infrared and lidar, and combines spatiotemporal graph convolutional network (ST-GCN) to model target interaction relationships, thereby improving the stability of target detection and accurately identifying abnormal group events.
In harsh environments, the target detection stability rate has been improved to 92%, and the accuracy of identifying abnormal group events has reached 89.7%, significantly improving the proactive early warning capability and response efficiency of the security system. It has been successfully applied to intelligent inspection of transportation hubs and perimeter protection of key areas.
Smart Images
Figure CN121884504A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of artificial intelligence recognition technology, and in particular relates to an intelligent monitoring system and its control method. Background Technology
[0002] The public safety sector is currently facing a prominent contradiction between the explosive growth of surveillance data and low security efficiency. Statistics show that over 25 billion hours of surveillance video are generated globally every day. However, traditional monitoring methods relying on manual patrols suffer from inherent flaws such as slow response times, high missed detection rates, and limited analytical dimensions, revealing serious shortcomings in critical scenarios like counter-terrorism early warning and emergency response. This situation urgently necessitates a shift in the monitoring paradigm from "passive recording" to "active perception" through intelligent technologies. In recent years, artificial intelligence technologies, represented by deep learning, have provided crucial support for this transformation—the mAP (memory accuracy) index of object detection algorithms has improved from 63.3% for AlexNet in 2012 to 89.2% for YOLOv8, and the IDF1 score of multi-object tracking (MOT) algorithms has exceeded 85%, making real-time and accurate monitoring in complex scenarios possible.
[0003] However, existing intelligent monitoring systems still face three major technical bottlenecks: First, the recognition performance of a single visual modality drops sharply in complex environments such as low light and rain / fog, with actual tests showing that the recall rate for pedestrian detection at night is generally below 60%; second, traditional detection algorithms struggle to understand the spatiotemporal relationships between targets, resulting in an accuracy rate of less than 75% for identifying abnormal group behavior; and third, centralized cloud computing architectures suffer from problems such as high network latency and the risk of data privacy leaks. Summary of the Invention
[0004] The technical problem of this invention is: Existing intelligent monitoring systems still face three major technical bottlenecks: First, the recognition performance of a single visual modality drops sharply in complex environments such as low light and rain / fog, with actual tests showing that the recall rate for pedestrian detection at night is generally below 60%; second, traditional detection algorithms struggle to understand the spatiotemporal relationships between targets, resulting in an accuracy rate of less than 75% for identifying abnormal group behavior; and third, centralized cloud computing architectures suffer from problems such as high network latency and the risk of data privacy leaks.
[0005] The purpose of this invention is: This invention aims to construct a distributed intelligent monitoring architecture based on multimodal feature fusion. By complementing multi-source data from visible light, infrared, and lidar, the target detection stability in harsh environments is improved to over 92%. Furthermore, by employing a spatiotemporal graph convolutional network (ST-GCN) to model target interaction relationships, the accuracy of identifying abnormal group events reaches 89.7%. This technology has been successfully applied to several typical scenarios, such as intelligent inspection of transportation hubs and perimeter protection of key areas, significantly improving the proactive early warning capabilities and response efficiency of security systems.
[0006] The technical solution of this invention is: On one hand, the present invention proposes an intelligent monitoring system, comprising: a lidar detection component (1), an optical detection component (2), a control component (3), a mounting carrier (4), a fixed base (5), a display screen (6), a transmission triggering device (7), and a transmission device (8), wherein: the lidar detection component (1) is disposed above the optical detection component (2), and the lidar detection component (1) is used to detect the actual position coordinates of targets in normal environments and in complex environments such as low illumination and rain / fog with a single visual modality; the optical detection component (2) includes two types of optical detection cameras, visible light and infrared, and the detection fields of the two optical detection cameras are adjusted to overlap. The visible light image and infrared image of the detected intruder are superimposed to facilitate the precise fusion of the optical information feature data of visible light and infrared light. The control component (3) is set on the mounting carrier (4). The control component (3) is electrically connected to the optical detection component (2) and the lidar detection component (1) respectively. It is used to accurately determine the category attributes and coordinate information of the target based on the actual position coordinates of the target under different environments and the precisely fused optical information feature data. At the same time, the identification information of the intruder is displayed on the display screen (6), and the on / off state of the transmission trigger device (7) is controlled accordingly. The alarm is executed, and the locking or transmission state of the transmission device (8) is controlled.
[0007] Preferably, a lidar detection component (1) is disposed above an optical detection component (2), and the lidar detection component (1) is used to detect the actual position coordinates of the target.
[0008] Preferably, the optical detection component (2) is disposed on both sides of the control component (3). The optical detection component (2) includes a visible light detection camera (2-1) and an infrared optical detection camera (2-2). The detection areas of the two detection components overlap. The optical detection component detects and identifies the category, size and other attributes of the target object based on the optical information of the target object.
[0009] Preferably, the detection fields of the visible light detection camera (2-1), the infrared optical detection camera (2-2), and the lidar detection component (1) are highly overlapping. The overlapping area of the detection fields of the three sets of detection devices is the effective detection area of the intelligent monitoring system. Preferably, the optical detection assembly (2) has a housing that protects the camera device. The housing is designed as a split housing, consisting of a detachable top cover and a bottom plate. A cavity is formed between the top cover and the bottom plate, and the optical detection assembly (2) is placed inside the cavity.
[0010] Preferably, the outer surface of the housing cover is designed with heat dissipation fins to increase the heat dissipation area and maintain overall sealing; the bottom plate is designed with a tortuous airflow path to prevent water from directly entering, while allowing heat to escape slowly.
[0011] Preferably, the control component (3) is disposed inside the mounting carrier (4), and the control component (3) is electrically connected to the optical detection component (2) and the lidar detection component (1) respectively, for determining the category attributes and final coordinates of the target object based on the optical information features and the lidar detection position.
[0012] Preferably, the control component (3) controls the optical detection component (2) and the lidar detection component (1) to start working respectively. The optical detection component detects the optical information features of the target object, and the lidar detection component (1) detects the actual position coordinates of the target object. The control component (3) receives visible light, infrared image data and laser echo signals respectively, and analyzes the target feature information. The control component (3) performs target association based on the optical feature information and laser echo signal features to determine the category attributes of the target object and the final actual coordinates. The control component (3) receives control commands and performs multimodal information fusion on the detection results. By analyzing visible light, infrared and laser echo data, it makes up for the limitations of a single sensor, performs cross-verification with multiple sensors, reduces the false detection rate, and enables the system to maintain its function even when some sensors fail.
[0013] On the other hand, this invention proposes an intelligent monitoring method that utilizes the aforementioned intelligent monitoring system for monitoring. This method includes the following steps: The lidar detection component (1) detects the actual position coordinates of targets in normal environments and in complex environments such as low light and rain / fog using a single visual modality; The optical detection component (2) detects overlapping visible light and infrared images of the intruding target and accurately fuses the optical information feature data of visible light and infrared light. The control component (3) accurately determines the category attributes and coordinate information of the target based on the actual position coordinates of the target in different environments and the precisely fused optical information feature data. At the same time, it displays the identification information of the intruding target on the display screen (6) and controls the on / off state of the launch trigger device (7) accordingly, and controls the execution of alarms, and controls the locking or launching state of the launch device (8).
[0014] The advantages and beneficial effects of this invention are: This invention constructs a distributed intelligent monitoring architecture based on multimodal feature fusion. By complementing multi-source data from visible light, infrared, and lidar, the target detection stability rate in harsh environments is improved to over 92%. Furthermore, by employing a spatiotemporal graph convolutional network (ST-GCN) to model target interaction relationships, the accuracy of identifying abnormal group events reaches 89.7%. This technology has been successfully applied to several typical scenarios, such as intelligent inspection of transportation hubs and perimeter protection of key areas, significantly improving the proactive early warning capabilities and response efficiency of security systems. Attached Figure Description
[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 This invention provides a schematic diagram of the structure of an intelligent monitoring system according to one embodiment. Figure 2 This invention provides a schematic diagram of the structure of an intelligent monitoring system according to another embodiment of the present application. Figure 3 A connection diagram of the system is shown; Reference numerals: LiDAR detection component (1), visible light detection camera device (2-1), infrared optical detection camera device (2-2), control component (3), mounting vehicle (4), fixed base (5), display screen (6), transmission trigger device (7), and transmission device (8). Detailed Implementation
[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0018] It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other, and the various embodiments can be referenced and cited in each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0019] See Figures 1 to 3This application provides a target range intelligent monitoring system based on multi-sensor information fusion, comprising: a lidar detection component (1), an optical detection component (2), a control component (3), a mounting carrier (4), a fixed base (5), a display screen (6), a launch trigger device (7), and a launch device (8). The optical detection component includes a visible light detection camera (2-1) and an infrared optical detection camera (2-2). The visible light detection camera (2-1) is used to detect the visible light information reflected by the target object, and the infrared optical detection camera (2-2) is used to detect the infrared optical information of the target object, and their detection fields of view overlap. The lidar detection device (1-1) is disposed in the protective housing (1-2) and is used to detect the position information of the target. The control component (3) is installed in the integrated mounting carrier (4) and is internally electrically connected to the lidar detector (1-1), the visible light detection camera (2-1), and the infrared optical detection device (2-2). The mounting vehicle (4) is fixed on the fixed base (5), and the photoradar detection component (1), optical detection component (2), and control component (3) can perform a 360-degree circumferential scan to obtain visible light images, infrared images, and lidar image information of the entire target range, which can be displayed separately or simultaneously. The control component (3) fuses the multiple detection information to obtain the type and location information of the intruding target, and sends the control information to the launch trigger device (7) based on this criterion, thereby controlling whether the launch device (8) should cease fire in an emergency.
[0020] Example 1 This application provides an intelligent monitoring system for a live-fire test range. It determines whether a dynamic target exists within the monitoring detection range (the range of the bullet impact point). Upon determining the presence of a dynamic target, it immediately controls the power-off of the launch trigger device (7), thereby ensuring that the launch device cannot power on and fire the test subject. This device includes: a laser radar detection component (1), an optical detection component (2), a control component (3), a mounting carrier (4), a fixed base (5), a display screen (6), a launch trigger device (7), and a launch device (8).
[0021] The optical detection components (2) are set on both sides of the mounting carrier (4). The optical detection components (2) include a visible light detection camera (2-1) and an infrared optical detection camera (2-1). The detection areas of the two sets of detection components overlap. The optical detection components detect and identify the category, size and other attributes of the target object based on the optical information of the target object.
[0022] In this embodiment of the application, the intelligent monitoring system consists of two sets of camera devices and a lidar detection component (1) forming a multimodal detection array. By analyzing visible light, infrared and laser echo data, it makes up for the limitations of a single sensor, cross-verifies with multiple sensors, reduces the false detection rate, and enables the system to maintain its function even when some sensors fail.
[0023] Furthermore, the intelligent monitoring system integrates visible light, infrared, and lidar data, combined with image stitching and deep learning technologies, to achieve broader coverage, longer-range detection, more accurate identification, and faster positioning of typical targets in complex environments. The system efficiently detects, identifies, and accurately locates sensitive targets such as drones and personnel, overcoming environmental interference and providing reliable, multi-dimensional information support for the safety of the test site around the clock.
[0024] Addressing the core need for efficient identification and early warning of intrusion targets in the security field, this system effectively solves the pain points of traditional monitoring such as "unclear visibility, numerous false alarms, and slow response" by breaking through key technologies such as real-time fusion of multi-source data (visible light + infrared + lidar), intelligent target analysis in complex environments (day and night, rain and fog), and precise delineation of electronic fences. It significantly improves the security protection level and proactive early warning capability of private territory.
[0025] See some possible implementations. Figure 1 and Figure 2 As shown, the visible light camera (2-1) and the infrared camera (2-2) include sensor carriers and cameras mounted on the carriers. The camera devices are installed in a protective housing, electrically connected to the central control component (3), and fixed on both sides of the mounting carrier (4). The detection field of view heights of the visible light camera (2-1) and the external camera (2-2) coincide. The lidar detection component (1) is electrically connected to the control component (3) and installed above the mounting carrier (4). The detection field of view heights of the lidar detection component (1) coincide with those of the visible light camera (2-1) and the infrared camera (2-2). The control component (3) is connected and fixed to the fixed base (5), through which the equipment is erected at a high point.
[0026] The optical detection component (2) consists of a visible light camera (2-1) and an infrared camera (2-2), used to collect visible light and infrared image information. The visible light and infrared images are aligned at the pixel level through spatiotemporal registration. The high-resolution texture / color features of visible light and the thermal radiation features of infrared are extracted by a dual-path deep learning network. Adaptive weighted fusion is performed at the feature level to enhance the salience of the target in complex environments such as day and night, rain and fog, and strong backlight. The fused features are input into the target detection network for joint inference. The target attributes are accurately classified using visible light detail information. At the same time, the infrared thermal features penetrate shielding and resist light interference, significantly improving the detection rate of intrusion targets and reducing false alarms, ultimately achieving all-weather robust recognition. The target is identified and classified according to the established dataset. Combined with the lidar detection component (1), the target is associated with optical feature information and laser echo signal features to determine the category attributes and final actual coordinates of the target.
[0027] The launch trigger device (7) has a network interface and can be connected to it via a network cable. The launch trigger device (7) determines the enabling state of the launch device (8) based on the control component (3)'s determination of whether there are dynamic targets or heat targets such as outsiders, vehicles, animals, etc. in the effective area. When it is determined that there are dynamic targets in the effective area, the 24V voltage of the launch trigger device (7) is disconnected, so that it cannot supply power to the product's electric primer and cannot complete the shooting action, thereby eliminating the safety risk of outsiders or vehicles entering the target lane during shooting.
[0028] System architecture design: The entire system's workflow can be clearly divided into the following stages: 1.1.1 First Phase: System Initialization 1. Configuration loading: (1) Create a security controller instance, which can optionally load the JSON configuration file. (2) Set default parameters: detection interval (0.1 seconds), safety delay (2 seconds), minimum target size, etc. (3) Initialize the system status to SAFE (safe state), and the voltage supply is normal. 2. Component Initialization (1) Set up a logging system to record running status and error information. (2) Initialize the OpenCV background subtractor for motion detection. (3) Prepare the YOLO target classifier (optional function) (4) Initialize the target tracking dictionary and region configuration list 1.1.2 Second Phase: Regional Configuration 1. Definition of Safe Zones (1) Users define danger zones (polygon coordinates) using add_danger_zone(). (2) Users define safe zones using add_safe_zone() (3) The system uses the ray casting algorithm to determine whether the target is within the dangerous polygon. 1.1.3 Third Stage: Startup Detection (Multi-threaded Execution) 1.1.41. Start the detection thread Call start_detection() to start two parallel threads: 2. Detection thread: (1) Open the camera and acquire the video stream at the configured resolution. (2) Perform moving target detection once every 0.1 seconds: 1) Apply background subtraction to separate moving objects in the foreground. 2) Morphological operations to remove noise 3) Find the outline and filter out small targets. 4) Classify objects (people, vehicles, animals, etc.) based on aspect ratio and area. 3. Monitoring thread: (1) Continuously assess the safety status at a frequency of 20 Hz. (2) Real-time monitoring of the relationship between the target location and the danger zone 1.1.5 Fourth Stage: Target Tracking and Status Management 1. Target Tracking Process Assign a unique ID to the newly detected target Update existing targets by matching location distance. Remove expired targets that have not been updated for more than 5 seconds. Record the type, location, confidence level, and timestamp of each target. 2. Safety Status Assessment (1) Continuously check whether all tracked targets have entered dangerous areas. (2) State transition logic: 1) Safety → Danger: When a target enters the danger zone for the first time 2) Danger → Safety: After all targets leave the danger zone and a 2-second safety delay has elapsed... 1.1.6 Fifth Stage: Implementation of Security Controls 1. Voltage control (1) Entering a dangerous state: Immediately disconnect the 24V voltage and record a warning in the log. (2) Restoring to a safe state: After a safety delay, power supply is restored. (3) Emergency stop: Manually triggered, immediately cut off the voltage and lock the system. 2. Notifications and Callbacks (1) Triggering the state callback function when the state changes (2) Trigger the target callback function when a new target is detected. (3) All events are recorded to the log file in real time. 1.1.7 Stage Six: System Shutdown 1. Stop (1) Call stop_detection() to stop all threads (2) Release camera resources (3) Record system stop logs In some embodiments, the network program (only a portion is shown) may look like this: System architecture design: import cv2 import time import threading import numpy as np from enum import Enum from dataclasses import dataclass from typing import Callable, List, Tuple import logging import json class SystemStatus(Enum): SAFE = "safe" # Safe state, no dynamic targets DANGER = "danger"# Dangerous status, dynamic target detected. EMERGENCY = "emergency" # Emergency status # Object detection related self.detected_targets = {} self.next_target_id = 1 self.detection_enabled = True # Regional Configuration self.danger_zones = [] # List of dangerous zone coordinates self.safe_zones = [] # List of safe zone coordinates # Configuration parameters self.config = { "min_confidence": 0.5, "min_target_size": 1000, # Minimum target pixel area "detection_interval": 0.1, # Detection interval (seconds) "safety_delay": 2.0, # Safety delay (seconds) "max_target_age": 5.0, # Maximum target survival time "voltage_control_pin": 18,# GPIO control pin "camera_index": 0, "resolution": (640, 480) } # Load configuration # Thread control # Callback function Configure the logging system. """Set up the target detector""" # Motion detection using OpenCV's BackgroundSubtractor # Load the YOLO model for object classification (optional) self.setup_yolo_detector() self.logger.info("Target detector initialization complete") self.logger.error(f"Object detector initialization failed: {e}") "Set up the YOLO target classifier (optional enhancement)" # Here you can load a YOLO model to identify specific target types. # Since YOLO requires model files, we will use motion detection as the basis for now. self.use_yolo = False Load configuration file Add danger zones self.danger_zones.append(points) self.logger.info(f"Dangerous areas have been added: {points}") def add_safe_zone(self, points): Add a safe zone self.safe_zones.append(points) self.logger.info(f"Safe zone has been added: {points}") def is_in_danger_zone(self, position): "Check if the target is within the danger zone." if not self.danger_zones: return True # If no danger zone is defined, the entire screen is a danger zone. """Determine if a point is inside a polygon"""" …… def detect_motion_targets(self, frame): "Detecting moving targets" # Application Background Subtraction # Noise Removal # Find Outlines # Filtering small goals # Get bounding box # Classification Target Types (Simplified Version) """Classification Target Types (Simplified Implementation)""" w, h = size aspect_ratio = w / h # Simple classification based on aspect ratio and size if aspect_ratio>1.5 and w * h>5000: return TargetType.VEHICLE elif 0.5<= aspect_ratio<= 1.5 and w * h>2000: return TargetType.PERSON elif w * h<3000: return TargetType.ANIMAL else: return TargetType.UNKNOWN def update_targets(self, new_targets): """Update target tracking""" current_time = time.time() updated_targets = {} # Remove expired targets # Match and update existing targets # Calculate location distance # Update existing goals # Create a new target # Trigger new target callback # Check if the target has left "Check if the target has disappeared." # All targets have disappeared, initiate a safety delay. "Safety conditions met" "Assess the security status" # Record dangerous targets It should be noted that the above process operations can be combined to varying degrees. For the sake of brevity, the implementation methods of various combinations will not be elaborated here. Those skilled in the art can flexibly adjust the order of the above operation steps or flexibly combine the above steps according to actual needs.
[0029] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should be covered within the protection scope of the present invention.
Claims
1. An intelligent monitoring system, characterized by, include: The components include: a lidar detection assembly (1), an optical detection assembly (2), a control assembly (3), a mounting vehicle (4), a mounting base (5), a display screen (6), a transmission triggering device (7), and a transmission device (8), wherein: The lidar detection component (1) is disposed above the optical detection component (2). The lidar detection component (1) is used to detect the actual position coordinates of the target in a normal environment and in a complex environment with a single visual modality, such as low illumination, rain, and fog. The optical detection component (2) includes two types of optical detection cameras: visible light and infrared light. The detection fields of the two types of optical detection cameras are adjusted to overlap, so that the visible light image and infrared image of the detected intruding target overlap, so as to facilitate the accurate fusion of the optical information feature data of visible light and infrared light. The control component (3) is mounted on the mounting carrier (4). The control component (3) is electrically connected to the optical detection component (2) and the lidar detection component (1) respectively. It is used to accurately determine the category attributes and coordinate information of the target based on the actual position coordinates of the target in different environments and the precisely fused optical information feature data. At the same time, it displays the identification information of the intruding target on the display screen (6) and controls the on / off state of the launch trigger device (7) accordingly, and controls the execution of alarms, controls the locking or launching state of the launch device (8).
2. The intelligent monitoring system of claim 1, wherein, in: The lidar detection component (1) is installed above the mounting carrier (4) to prevent the optical detection component (2) from blocking it, and to adjust its precise relative position with the optical axis of the optical detection component (2).
3. The intelligent monitoring system of claim 1, wherein, in: The optical detection component (2) includes a visible light detection camera (2-1) and an infrared optical detection camera (2-2). The two optical detection devices are respectively installed on both sides of the control component (3) and fixed on the mounting carrier (4). Their detection fields of view are adjusted to overlap so that the visible light image and infrared image of the intruding target are accurately overlapped.
4. The intelligent monitoring system of claim 1, wherein, in: The control component (3) is electrically connected to the infrared detection component (2-2), the visible light detection component (2-1), and the lidar detection component (1) within the mounting carrier (4); The control component (3) controls the infrared detection component (2-2), the visible light detection component (2-1), and the lidar detection component (1) in real time, while acquiring their detection information and performing information fusion to obtain the classification information and location information of the intruding target for subsequent decision-making.
5. The intelligent monitoring system according to claim 1, characterized in that, in: The mounting carrier (4) is used to carry the infrared detection component (2-2), the visible light detection component (2-1), the lidar detection component (1), and the control component (3). It is manufactured by integral casting, and each component is distributed around the control component (3) to ensure the relative position and installation accuracy of each detection component.
6. The intelligent monitoring system according to claim 1, characterized in that, in: The infrared detection component (2-2), visible light detection component (2-1), lidar detection component (1), and control component (3) are mounted on the fixed base (5) via the mounting carrier (4); The fixed base (5) is a hollow rotating structure that can drive the above-mentioned detection component carrier to perform a 360-degree scan of the monitoring area, thereby obtaining a circumferential visible light image, an infrared image, a lidar image of the monitoring area, and a fused image of the three images, which can be displayed on the screen.
7. The intelligent monitoring system according to claim 1, characterized in that, in: The launch triggering device (7) has a network interface and a satellite communication interface. The launch triggering device (7) determines the enabling state of the launch device (8) according to the instructions of the control component (3).
8. The intelligent monitoring system according to claim 1, characterized in that, in: The optical detection component (2) acquires visible light and infrared image information through a visible light camera (2-1) and an infrared camera (2-2). It achieves pixel-level alignment of visible light and infrared images through spatiotemporal registration. It also extracts high-resolution texture / color features of visible light and thermal radiation features of infrared light through a dual-path deep learning network. Adaptive weighted fusion is performed at the feature level to enhance the salience of the target in complex environments such as day and night, rain and fog, and strong backlight. The target detection network integrates feature inputs for joint inference, accurately classifying target attributes using visible light detail information. Simultaneously, it leverages the penetration and light interference resistance of infrared thermal features, significantly improving the detection rate of intrusive targets and reducing false alarms, ultimately achieving robust all-weather recognition. Based on the established dataset, the target is identified and classified, and combined with the lidar detection component (1), target association is performed based on optical feature information and laser echo signal characteristics to determine the target's category attributes and final actual coordinates.
9. The intelligent monitoring system according to claims 1-8, characterized in that, in: The launch trigger device (7) determines the enabling state of the launch device (8) based on the control component (3)'s determination of whether there are dynamic targets or heat targets such as outsiders, vehicles, animals, etc. within the effective area. When it is determined that there are dynamic targets within the effective area, the 24V voltage of the launch trigger device (7) is disconnected, so that it cannot supply power to the product's electric primer and cannot complete the shooting action, thereby eliminating the safety risk of outsiders or vehicles entering the target lane during shooting.
10. An intelligent monitoring method, using an intelligent monitoring system according to any one of claims 1-9 for monitoring, characterized in that, The method includes the following steps: The lidar detection component (1) detects the actual position coordinates of the target in a normal environment and in a complex environment with a single visual modality, such as low light and rain / fog. The optical detection component (2) detects overlapping visible light and infrared images of the intruding target and accurately fuses the optical information feature data of visible light and infrared light. The control component (3) accurately determines the category attributes and coordinate information of the target based on the actual position coordinates of the target in different environments and the precisely fused optical information feature data. At the same time, it displays the identification information of the intruding target on the display screen (6) and controls the on / off state of the launch trigger device (7) accordingly, and controls the execution of alarms, and controls the locking or launching state of the launch device (8).