CN-122020542-A - Urban forest resource monitoring system and monitoring method based on multi-source remote sensing fusion

CN122020542ACN 122020542 ACN122020542 ACN 122020542ACN-122020542-A

Abstract

The invention relates to a multi-source remote sensing fused urban forest resource monitoring system and method, which construct a dual-core system architecture for sound and vibration cooperative control and edge intelligent processing, collect remote sensing data of urban complex operation environments with high quality, realize full-element centimeter-level sensing and robust obstacle avoidance of the urban forest complex environments, and guarantee operation safety and data integrity. Based on the interpretable AI and the multi-mode fusion knowledge acquisition, the scale pushing of the single wood to the stand is implemented, the high-precision intelligent analysis of stand scale parameters is completed, and the automatic and quantitative inversion of stand scale key parameters is realized at the airborne edge. The system automatically calculates key stand structural factors such as stand density, closing degree, average height, variation coefficient and the like by aggregating single wood detection results, realizes single machine intelligence and group coordination, and enables large-scale and high-efficiency urban forest resource investigation and dynamic monitoring.

Inventors

SHANG KANKAN
ZHANG GUOWEI
RUAN JUNJIE
ZHANG QIAN
WANG QING
LUO PEIWEN

Assignees

上海辰山植物园

Dates

Publication Date: 20260512
Application Date: 20260129

Claims (10)

1. The urban forest resource monitoring system is applied to a low-altitude flight platform and is characterized by comprising a collaborative sensing acquisition unit, a sound and vibration collaborative control unit, a fusion positioning synchronization unit, a time synchronization management unit, an edge intelligent processing unit and a full-link data management unit, wherein the units are connected with each other; The collaborative awareness acquisition unit is used for acquiring multi-mode original data and original observation values; The sound and vibration cooperative control unit is used for controlling the radiation noise in a first preset range of the flying platform within a first preset numerical range and controlling the noise acceleration root mean square error of the platform body within a second preset numerical range; The fusion positioning synchronization unit is used for receiving the multi-mode original data and the original observed value, executing fusion positioning calculation and generating a high-precision carrier pose; the time synchronization management unit is used for distributing unified clock signals to each sensor in the collaborative awareness acquisition unit, distributing time stamps to all the multi-mode original data and performing time alignment on the multi-mode original data; The edge intelligent processing unit comprises a first processing channel and a second processing channel, wherein obstacle detection and obstacle avoidance path planning are carried out through the first processing channel, and crown detection, tree species identification, inversion of single wood structure parameters, stand scale parameter calculation and stand attribute division are carried out through the second processing channel on the multi-mode raw data and the carrier pose after time alignment to generate structured forest information; The full-link data management unit is used for receiving the structured forest information, compressing the structured forest information into a vector abstract, and packaging and encrypting the vector abstract, the privacy processing record, the task metadata and the time stamp to generate a traceable task compliance data packet.
2. The multi-source remote sensing fused urban forest resource monitoring system according to claim 1, wherein, The collaborative perception acquisition unit further comprises a vibration triggering type online calibration module, wherein the vibration triggering type online calibration module is configured to calculate a vibration acceleration root mean square value of the flying platform in real time, and correct the multi-mode original data and the original observed value in real time when the vibration acceleration root mean square value exceeds a preset threshold range.
3. The multi-source remote sensing fused urban forest resource monitoring system according to claim 1, wherein, In the second processing channel, the tree species identification includes: Constructing a multi-modal feature vector, wherein the multi-modal feature comprises basic features consisting of NIR reflectivity, NDVI index, red edge position, spectrum first derivative, crown size and projection area, multi-scale texture energy and crown sparsity index, and further fusion extracting auxiliary morphological features of branch angle directivity, thermal infrared gradient and crown morphology asymmetry when the quality of point cloud meets preset conditions; the forest stand attribute division comprises a forest stand age group division, and specifically comprises the following steps: Based on crown detection and tree species identification results, generating forest stand scale characteristic parameters in a target analysis area in a polymerization way, wherein the characteristic parameters comprise average tree height, standard deviation of tree height, average crown width, tree density, canopy density and average spectrum index of the forest stand; and obtaining the age group category to which the stand belongs according to the stand scale characteristic parameter.
4. The multi-source remote sensing fused urban forest resource monitoring system according to claim 1, wherein, In the second processing channel, the single-log structure parameter inversion includes an estimation of a single-log chest diameter; the forest stand scale parameter calculation comprises calculation of forest accumulation or biomass, and specifically comprises the following steps: The method comprises the steps of obtaining crown amplitude information based on crown detection, obtaining trunk point cloud based on laser radar three-dimensional point cloud, estimating a single wood breast diameter by combining tree species identification results and a ground three-dimensional green quantity model, calculating single wood timber or biomass according to tree species, estimated breast diameter and tree height of single wood, aggregating all single wood timber or biomass and spatial distribution thereof in a target analysis area, and calculating and outputting total accumulation or biomass of forests in the area and accumulation or biomass in unit area.
5. The urban forest resource monitoring system with multi-source remote sensing fusion according to claim 2, wherein the sound and vibration cooperative control unit comprises a low-sound-level aerodynamic component, a dynamic noise management component and an active vibration suppression component; the dynamic noise management component is used for adjusting the rotating speed of the rotor wing and the noise frequency spectrum in real time, and is matched with the low-sound-level aerodynamic component to control the vibration noise acceleration RMS of the platform within the range of 0.2-0.6 g; the active vibration suppression component is used for monitoring accelerometer signals of platform vibration in real time; and the vibration triggering type online calibration module is in signal connection with the active vibration suppression component and the collaborative sensing acquisition unit and executes calibration based on the accelerometer signals.
6. The multi-source remote sensing fused urban forest resource monitoring system of claim 5, wherein the system is configured to perform the following compliance strategy in conjunction with the cooperative control of sound vibrations in a flight mission: When the flying platform enters a preset noise sensitive area according to positioning information, a low-noise flying mode is automatically started, and the mode reduces the reference rotating speed of the rotor wing through the dynamic noise management component and optimizes a modulation strategy; The automatic execution of the collaborative compensation control comprises the steps of adjusting attitude loop parameters of a flight controller in a linkage manner to maintain the stability of a platform and optimally planning a low disturbance route, synchronously adjusting acquisition parameters of a multispectral/hyperspectral imaging component and/or a thermal imaging sensor, and for the thermal imaging sensor, the parameter adjustment comprises the steps of optimizing the frame integration time within a range of 1-10 milliseconds and adjusting the signal amplification gain within a range of 1-16 times so as to realize the optimal sensitivity and the dynamic range of a target temperature difference under the dynamic flight condition, and automatically associating and marking the time period data as low noise mode-vibration compensated data by a system if the vibration triggering type online calibration module is activated.
7. A method for monitoring urban forest resources by multi-source remote sensing fusion, which is applied to the urban forest resource monitoring system by multi-source remote sensing fusion as claimed in any one of claims 1 to 6, and is characterized by comprising the following steps: s1, acquiring multi-mode original data and acquiring an original observation value by a dual-antenna carrier phase difference global navigation satellite system receiver; s2, fusion positioning and time alignment are carried out by using the acquired data; s3, carrying out airborne edge parallel processing and real-time compliance; s4, data minimization and trusted encapsulation; and S5, carrying out multimode self-adaptive feedback and verification, supporting a communication link of 5G/4G, wi-Fi, beidou short message and LoRa, selecting or switching paths according to the quality of the real-time link, returning the task compliance data packet, and writing a key hash value into a time stamp chain or a block chain for verification.
8. The method of claim 7, wherein S1 comprises: s11, according to the flying environment noise background and a preset sensitive area map, automatically selecting and switching to a corresponding noise control gear by a dynamic noise management component, and adjusting the modulation frequency of the electronic speed regulator to be in a range of 300-800 Hz; s12, before multispectral/hyperspectral or thermal imaging data are collected, the active vibration suppression component starts pre-detection, and a baseline vibration value is recorded; And S13, continuously comparing the real-time vibration value with the threshold value in the data acquisition process, starting a calibration sub-process within 10-30ms once triggered, performing on-site radiometric calibration and atmospheric correction, and calling a preset spectral response function or a temperature-pixel deviation matrix to correct the current and subsequent frames of data.
9. The method of claim 7, wherein S3 comprises: S31, fusing vision, laser point cloud and millimeter wave radar data, performing multi-mode sensing at a frequency of 20-50Hz to obtain obstacle position, size, relative speed and reliability score (0-1), re-planning a track at a frequency of 10-20Hz to ensure that the unmanned plane keeps a lateral obstacle avoidance distance of 0.5-1.5m, and keeping a dynamic obstacle response time delay of <100ms; S32, intelligently extracting forest information, namely performing crown instance segmentation and single wood positioning by combining carrier pose based on RGB images and optional multispectral/hyperspectral/thermal imaging data, performing multi-mode fusion by adopting features based on the top 10 of the SHAP value importance ranking to finish tree classification, and inverting single wood structure parameters, wherein the single wood structure parameters at least comprise tree height, crown amplitude and estimated breast diameter; s33, calculating stand scale parameters and dividing attributes, namely carrying out statistics and space analysis on a preset or dynamically defined stand analysis unit based on the single tree species, the position, the structure parameters, the volume and the biomass obtained in the step S32; and S34, performing real-time privacy stripping, namely running a detection algorithm of a key infrastructure target in real time for each frame of visual image in the processing procedures of the steps S31, S32 and S33, and finishing blurring or geographic information security processing within 30-50 ms.
10. The method according to claim 9, wherein the method further comprises the steps of preloading the sound vibration control strategy according to the task area attribute in the flight path planning stage, and pre-reducing the planning speed of a leg required to traverse the noise sensitive area and marking the leg as first labeling information; in the task execution process, if the system records and displays that the calibration times of a certain navigation segment triggered by continuous vibration exceeds a preset value, automatically adding second labeling information into a data compliance packet generated at the rear end; The method further comprises the steps of dynamic compliance operation, namely automatically loading a flight restriction area and privacy rules issued by a city management department, performing risk assessment at the frequency of 2-5Hz, triggering a compliance degradation mode when entering a sensitive area, automatically planning a low-disturbance route, maintaining a lateral safety margin of 0.5-1.2m, limiting the maximum navigational speed, supporting multi-machine cooperation, ensuring no repetition of coverage through a task synchronization protocol of 0.5-2Hz, and performing space-time consistency reconstruction after GNSS signals are lost for a short time.

Description

Urban forest resource monitoring system and monitoring method based on multi-source remote sensing fusion Technical Field The application relates to the technical field of forest resource checking and monitoring, in particular to a multi-source remote sensing fusion urban forest resource monitoring system and method. Background With the acceleration of urban ecological civilization construction, urban forests are used as important green basic facilities, and the demands of resource census, health diagnosis, carbon sink metering and fine management are increasingly urgent. The traditional ground investigation method relying on manpower has the defects of low efficiency, limited coverage, strong subjectivity and the like, and is difficult to meet the requirements of modernization, high frequency and large-scale monitoring. Therefore, the low-altitude remote sensing technology using unmanned aerial vehicles as carriers is becoming a mainstream means for urban forest resource monitoring. However, when the unmanned aerial vehicle remote sensing technology is applied to a ground structure complex area in a urbanization area, a series of systematic technical challenges which are not encountered in traditional forestry or open-land operation are faced, and the prior art scheme has significant disadvantages, which are specifically expressed in the following aspects: 1. Insufficient environment sensing and safety obstacle avoidance capability, and difficulty in coping with urban three-dimensional space obstacle The urban forest area environment is highly complex, and a large number of obstacles which are naturally and artificially interwoven exist, namely, on one hand, the tree canopy is various in form and dense in branches and leaves, and can shield the sensor, and on the other hand, the urban forest area environment is covered with obstacles with weak dynamic or reflective characteristics such as wires (1-5 cm thick), cables, glass curtain walls, street lamps, pedestrian vehicles and the like. The existing forestry monitoring unmanned plane mostly adopts a single vision or lightweight radar scheme, the perception capability is rapidly reduced under low light, haze and backlight conditions, the effective detection distance of the slender linear obstacle is short (usually less than 5 m), and the omission ratio is high. The unmanned aerial vehicle cannot fly safely at low altitude and short distance in a high-density town area, the capability of acquiring high-resolution canopy texture and structure data is severely restricted, and the contradiction between flight safety and observation precision is outstanding. 2. Poor consistency of space-time reference, and influences data precision and long-term comparability The urban architecture and the dense forest crown cause serious shielding and multipath effects on the global navigation satellite system signals, so that the positioning signals are frequently interrupted or the accuracy is deteriorated. The existing unmanned aerial vehicle system generally adopts a loose coupling scheme of a consumer-level GNSS module and an inertial measurement unit, positioning errors are rapidly accumulated after signals are lost, and drift rate can reach several meters per minute. Meanwhile, the time synchronization mechanism of a hardware level is lacked among airborne multisensors (visible light, multispectral, hyperspectral and laser radar), and the data timestamp alignment error is often more than millisecond level. The inconsistency of the space-time references makes the data acquired by different frames and different sensors difficult to accurately register and fuse, and seriously influences the accuracy of quantitative analysis such as tree height inversion, crown amplitude measurement, change detection and the like, and the comparability of long-term monitoring data. 3. Data privacy and compliance risk are outstanding, and source management mechanism is lacked Urban forest monitoring inevitably captures images containing sensitive information such as private yards, communication stations, etc. The current common practice is to completely transmit the original image or video stream back to the ground station or cloud for processing, which is to spread the original data containing a large amount of personal information in uncontrollable transmission and storage links, and there is a huge risk of privacy disclosure. Although some schemes attempt to perform blurring processing at the back end, the mode of "original data out of machine" violates the data minimization principle, and can not provide verifiable proof of privacy processing, and also cannot meet compliance requirements of laws and regulations such as "personal information protection law". 4. The data processing paradigm is behind, and the real-time performance and bandwidth contradiction are obvious The existing system highly depends on the computing power of a ground station or a cloud server to carry out heavy processi