156 results about "Adaptiv" patented technology

The invention relates to the technical field of radar data processing, in particular to a method which is suitable for a multi-external radiation source and multi-station system, carries out single-frame and two-frame multi-target measurement value matching under the condition of no bistatic distance information, has good adaptability to the multi-target positioning problem of external radiation source systems of different systems, and is high in positioning accuracy. The bistatic joint multi-target positioning method based on the external radiation source system is beneficial to the processes of multi-target positioning, real-time tracking, track generation and the like, and comprises the following steps: matching multi-source and multi-station angle measurement values to obtain candidate association, and estimating the intersection point position and the speed state of the angle measurement values; approximately constructing a linear equation of bistatic speed and angle measurement value errors, establishing measurement value error estimation of candidate associations and a real target, and removing wrong candidate associations by using an association hypothesis decision; and establishing a cost function according to the measured value error estimation of the candidate association between the two frames and the state mahalanobis distance, and removing the error candidate association between the two frames by using an association hypothesis decision.

The invention relates to the technical field of intelligent inspection of power equipment, and particularly discloses a power transmission line inspection system based on an autonomous correction long-endurance unmanned airship. The system aims to solve the problems of short endurance, small operation radius, single image angle, intermittent operation, leak detection, low precision, poor environmental adaptability and the like caused by dependence on manpower or GPS navigation in the conventional unmanned aerial vehicle inspection. The system comprises an energy supply module, a data acquisition module, an intelligent control module and a communication and data return module. Wherein the energy supply module is integrated with a solar energy collection, storage and intelligent scheduling unit to realize day-and-night continuous energy supply; the intelligent control module adopts an RT-DETR visual algorithm to identify a tower in real time, performs automatic deviation correction by fusing pose information, and accurately tracks a preset track; the data acquisition module supports multi-angle high-definition imaging, and the communication module guarantees real-time data return. And high-precision unmanned inspection with ultra-long endurance, full-automatic and wide-area continuous coverage is integrally realized.

The invention belongs to the technical field of artificial intelligence, particularly relates to the technical field of reinforcement learning, and particularly relates to an unmanned system autonomy evaluation method based on reinforcement learning, which can be applied to intelligent unmanned aerial vehicle manufacturing, and comprises the following steps: constructing an unmanned aerial vehicle autonomy evaluation index system; constructing an unmanned system autonomy scoring model based on reinforcement learning; and training an autonomous scoring model of the unmanned system. According to the method, a three-level index system taking the autonomy of the unmanned aerial vehicle as a root node is established, a double neural network structure based on reinforcement learning is designed, and an autonomy score is calculated. A multi-head attention mechanism is embedded in the value network, and adaptive learning of each leaf index weight is realized. According to the method, the importance weights of different evaluation dimensions are dynamically adjusted, and accurate and adaptive autonomy scores are provided. According to the method, the problems of fixed weight and lack of adaptability in a traditional evaluation method are solved, and the evaluation strategy can be dynamically optimized according to actual task requirements.

The invention relates to a transformer substation bird three-dimensional positioning method and system based on double thermal infrared imagers, and belongs to the technical field of transformer substation safety monitoring. The system comprises a data acquisition module, a background heat source filtering module, a three-dimensional coordinate fusion module, a deep learning recognition module, a multi-modal meteorological compensation module, a trajectory prediction and early warning module and a data output and linkage control module. The method comprises the following steps: acquiring data cooperatively through double thermal infrared imagers, accurately identifying a bird target by using a deep learning model after background filtering and temperature threshold preliminary screening, calculating a three-dimensional space position of the bird target through coordinate fusion, and dynamically optimizing detection parameters in combination with real-time meteorological data. And finally, an improved Kalman filtering algorithm is adopted to predict a bird movement track, and early warning and linkage with a bird repelling device are realized. The problems of inaccurate positioning, poor environmental adaptability, early warning lag and the like in the prior art are solved, and the accuracy, the reliability and the active defense capability of bird monitoring of the transformer substation are improved.

The invention discloses an on-orbit automatic focusing method for a spaceborne space target observation camera, relates to the technical field of space application of spaceborne cameras, and solves the problems of poor real-time performance and insufficient unattended operation adaptability of a manual focusing technology in the existing focusing technology. And meanwhile, in the automatic focusing technology, the hardware structure is complex, the space environment adaptability is weak, satellite-borne platform resources are limited and difficult to deploy, and focusing is difficult in a severe out-of-focus scene. According to the invention, the realization of automatic focusing is divided into a coarse adjustment stage and a fine adjustment stage. A large step pitch is adopted in the coarse tuning stage, rapid crossing is achieved, and an area adjacent to a positive focus position is rapidly positioned; in the fine adjustment stage, small step pitch is adopted, fine adjustment is achieved, and the peak position is accurately locked. The contradiction between the search speed and the positioning precision is balanced. When far away from the positive focus, a large-step-pitch fast approximation peak region is used, so that the number of iterations and the number of image acquisition frames are greatly reduced; when the focus is approached, small step pitch is switched for fine tuning, and the final precision is ensured.

The invention relates to the technical field of photoelectric detection and radars, in particular to a calibration method, system and device for a radar and a laser anti-collision device and a medium, and aims to track and lock the same dynamic non-cooperative target through the laser anti-collision device and the phased array radar. Synchronously acquiring spherical coordinate observation data of the dynamic targets in the public view field under respective coordinate systems; and after the spherical coordinates are converted into rectangular coordinates, calculating a point set centroid, constructing a covariance matrix and performing singular value decomposition based on a rigid body transformation model, and directly solving a rotation matrix and a translation vector from a radar coordinate system to a laser anti-equipment-free coordinate system, thereby completing calibration. No external absolute coordinate reference is needed, only relative observation data of equipment is utilized, and the method has the advantages of being rapid in deployment, high in environmental adaptability, low in cost and high in calibration speed. The problems that in the prior art, a calibration system depends on RTK or a high-precision map, the environmental adaptability is poor, and the deployment requirement is high are solved.

The invention provides an airborne infrared image ship detection method and system based on feature fusion and FVIM-HLOA, and relates to the technical field of target detection. According to the technical key points, the method comprises the following steps: obtaining an airborne infrared image containing a marine ship, and carrying out preprocessing on the airborne infrared image; extracting a multi-dimensional feature map from the preprocessed image; adopting a fuzzy variable index meta-heuristic algorithm to perform adaptive clustering on the multi-dimensional feature map to generate a region-of-interest mask; and obtaining an optimal segmentation threshold value by using an improved angular skill optimization algorithm, and carrying out multi-threshold adaptive optimal segmentation on the mask of the region of interest by using the optimal segmentation threshold value to obtain a final segmentation map. The method has the advantages of being high in feature discrimination capability, efficient and accurate in detection framework, high in optimization capability of an optimization algorithm, high in environmental adaptability and the like, and has good generalization capability and practical value.

The utility model relates to the technical field of automatic recharging, in particular to an automatic recharging system adopting a robot visual navigation and infrared receiving mode, which comprises a robot main body and a charging seat, and the robot main body comprises a power supply module, an infrared receiving module, a camera, an electric quantity detection module, a main control module, a driving module and a load. The charging seat comprises a charging circuit and an infrared transmitting module, and the infrared receiving module and the infrared transmitting module are matched with each other. According to the utility model, the robot visual navigation and the infrared receiving are combined, the advantages of the robot visual navigation and the infrared receiving are brought into full play, in an environment with good light, the camera collects surrounding environment images, and after the images are analyzed by the image processing unit, accurate position information is provided for the robot, and the robot is guided to move towards the charging seat; when light is poor or blocked, the infrared receiving module is matched with the infrared transmitting module of the charging base, it is ensured that the robot can stably receive signals, the environmental adaptability of the system is effectively enhanced, and the recharging success rate is increased.

The invention discloses a target self-adaptive positioning method and system for low-altitude routing inspection, and aims to solve the problems that target positioning depends on additional hardware, the complex environment adaptability is poor and the real-time performance is insufficient in the existing low-altitude routing inspection. According to the method, an unmanned aerial vehicle collects a target object image and extracts pixel coordinates, Beidou positioning data and IMU attitude information are fused, and normalized coordinates in a camera coordinate system are obtained through coordinate conversion; a dynamic virtual baseline technology is innovatively introduced, depth information of a target object is obtained through feature point tracking, motion solution and depth calculation, and finally three-dimensional longitude and latitude positioning of the target object is achieved by combining a rotation matrix and a coordinate conversion algorithm. The method does not need extra hardware, improves the positioning precision and real-time performance in a complex scene, and is suitable for the dynamic target positioning requirement in the field of low-altitude inspection.

The application belongs to the technical field of intelligent detection of building engineering, and particularly relates to an automatic defect identification system for fabricated buildings based on an unmanned aerial vehicle, which comprises a data acquisition module, a data registration module, a sleeve positioning module, a feature extraction module, a defect identification module and a report output module. The system collects multi-source heterogeneous data through the unmanned aerial vehicle, constructs a time-sequenced three-dimensional data field, eliminates the influence of environmental interference through pretreatment, extracts visible light image features, infrared thermal imaging features and three-dimensional point cloud geometric features in parallel, realizes comprehensive diagnosis of the internal and external filling state of grouting, accurately identifies insufficient hollows of grouting, late shrinkage hollows and defects of insufficient vertical depth of grouting, simultaneously carries out dynamic tracking, accurate quantification and tracing, effectively solves the problems that the existing detection of the filling state of sleeve grouting of fabricated buildings is single in dimension, poor in space-time continuity, weak in environmental adaptability and difficult to accurately quantify and trace defects, and improves the intelligent reliability of quality control.

The invention provides a robot dog multi-mode sensor fused substation equipment hidden danger detection device, and relates to the field of power equipment state monitoring. The hidden danger detection device for the substation equipment comprises a sensor integration module, an environment adaptation unit, a data fusion processing unit and a result output module. The sensor integration module fuses multi-mode data of a laser radar, a thermal infrared imager and the like; the environment adaptation unit guarantees stable operation in extreme weather; the data fusion processing unit realizes accurate hidden danger identification and risk grading through a CNN-LSTM algorithm; and the result output module supports real-time transmission and local backup. Through the arrangement of the plurality of sensors and the processing short circle, the problems of detection limitation and poor extreme environment adaptability of a single sensor are solved, hidden dangers of electrical, mechanical and material types can be comprehensively covered, the detection accuracy and the stability in extreme weather are remarkably improved, and reliable guarantee is provided for safe operation of substation equipment.

The invention provides a multi-modal data fusion unmanned aerial vehicle marine enteromorpha efficient monitoring method and system, and relates to the technical field of image processing, and the method comprises the steps: obtaining a natural optical image and an infrared image of marine enteromorpha collected by an unmanned aerial vehicle; inputting the natural optical image into a first backbone network for feature coding to obtain an optical modal initial feature map; inputting the infrared image into a second backbone network for feature coding to obtain an infrared modal initial feature map; performing cross-modal fusion and enhancement processing on the optical modal initial feature map and the infrared modal initial feature map to obtain an enhanced fusion feature map; and inputting the enhanced fusion feature map into a target detection network, and outputting position information, confidence and category information of the enteromorpha plaque. According to the method, the problems of poor adaptability and small target missing detection of single-mode monitoring in a complex environment are effectively solved, and the accuracy, the robustness and the early warning capability of enteromorpha monitoring are remarkably improved.

The invention discloses a self-adaptive fire early-stage detection method and system based on infrared, visible light and gas concentration information fusion. The method comprises the following steps: synchronously acquiring data through multiple sensors, extracting temperature, vision and gas characteristics, and obtaining environment context information; a three-level model including data-level association, feature-level complementation and decision-level adaptive weighted fusion is adopted for fusion, and the fusion weight is dynamically adjusted according to the environment context; and performing time domain analysis on the fusion result by using a time sequence confidence accumulation algorithm, and realizing hierarchical alarm based on risk integration. The system comprises a corresponding sensing layer, a data layer, a fusion decision-making layer and an application layer. According to the invention, extremely early and high-accuracy detection of flames and smoldering fire is realized, and the method has strong anti-interference capability and environmental adaptability.

The present application relates to the technical field of agricultural equipment, and discloses an alfalfa planting and topdressing device and a topdressing method, which comprise: a lightweight mobile frame, a connecting head connected with a power end, a front wheel and an infrared laser radar arranged at the front end, and an adjustable rear wheel group arranged at the rear end; a multi-modal detection unit integrated with the front wheel, a detection surface of the multi-modal detection unit and a scanning direction of the infrared laser radar form a spatial orthogonal layout, and the multi-modal detection unit comprises: a rolling trigger type gamma ray sensor which periodically emits rays while rotating with the front wheel; through the cooperative work of the multi-modal detection unit, the present device realizes comprehensive perception of soil deep nutrients, ion concentration, surface moisture and plant canopy, combines the terrain scanning function of the infrared laser radar, and dynamically generates a fertilizer formula and adjusts fertilization parameters, so that the precision and adaptability of fertilization are significantly improved, and the limitations of traditional single detection means are avoided.

The invention belongs to the field of image fusion, and discloses a parameter-free embedding remote sensing multi-modal image fusion method based on double-layer optimization, which comprises the following steps of: 1, generating a visible light basic convolution kernel and an infrared basic convolution kernel; 2, obtaining a visible light new convolution kernel and an infrared new convolution kernel; step 3, obtaining visible light features and infrared features; 4, semantic attention features are generated, and a final fusion image is output through decoding; step 5, the fusion knowledge is distilled to a Vision Transform sub-network, so that the Vision Transform sub-network is enabled to complete image fusion; according to the method, the convolution kernel is dynamically adjusted to adapt to different spatial resolutions and image sizes, so that the spatial adaptability is enhanced; recalibrating the position embedding vector by adopting a resolution alignment strategy so as to retain key mark features; various remote sensing data can be seamlessly processed without modifying a bottom network architecture, the completeness of the remote sensing data is kept, and multispectral data of different sensors can be compatible.

The invention relates to an electric power target hidden danger fine-grained identification method and system and a medium, and the method comprises the steps: employing a high-resolution camera and an infrared sensor carried by an unmanned plane to collect the visible light and infrared image data of an electric power line as a data set, and dividing the data set into a training set, a verification set and a test set; carrying out denoising, image enhancement and sample segmentation preprocessing operation on the collected image; detail features in the image are enhanced, and preparation is made for typical target detection under a complex distribution line background; through semantic-driven large model training, robust features of an interested target are automatically learned and extracted; performing fine-grained identification on the hidden danger of the power target in combination with the features after frequency domain scale transformation; and outputting an identification result. Through a frequency domain scale transformation technology, target features under different scales can be effectively processed, and the adaptability of the model to a complex environment is improved.

The application relates to the field of artificial intelligence and discloses a target recognition method and system based on deep learning, which comprises the following steps: acquiring visible light and infrared images of the same scene, generating illumination invariance features through an adversarial training network; modeling the feature sequence by using a space-time joint network to generate an initial recognition result; performing space-time verification on the initial recognition results of continuous multiple frames by using a Markov random field model to generate a calibrated recognition result; and performing online model calibration on the space-time joint network based on the correction deviation between the initial recognition result and the calibrated recognition result. The application also provides a recognition system for executing the method. Through the synergistic mechanism of multi-modal feature decoupling, space-time joint modeling and verification, and online feedback calibration, the application effectively improves the all-weather adaptability, recognition accuracy and trajectory continuity of the system in complex scenes such as illumination changes and target occlusions.

The application discloses a mobile fire detection system and method, and belongs to the technical field of fire detection. The system realizes over-distance accurate positioning and sub-pixel level precision. Through a unique "polynomial fitting + extreme value calculation" algorithm, the system can perform three times curve fitting on the signals of 2-3 detection units covered by the fire source imaging, and can analyze the sub-pixel level center coordinates by derivation. In combination with a preset coordinate mapping table, the fire source positioning precision can be improved to within 0.3 meters at a distance of 150 meters, and the specific tunnel post number and lane position can be accurately output, so that the limitation of the traditional detector that can only give an area alarm is completely changed. In terms of adaptability to complex environments, the application not only solves three major pain points of the traditional flame detector, i.e., positioning loss, weak anti-vibration and high false alarm rate, but also provides a complete set of high-reliability, intelligent and practical fire monitoring and emergency command tools for tunnel safety.

The utility model provides a kind of heating defroster of radar, including panel, heating wire, diaphragm layer, first metal sheet and second metal sheet. Current is transmitted to heating wire using first metal sheet and second metal sheet to generate heat. Heating wire is arranged on panel by diaphragm layer, panel is attached with radar or constitutes part of radar, so that the heat of heating wire is transmitted to panel, heating effect is realized, and the problem of fogging and frosting on the surface of radar can be effectively solved. At the same time, the second metal sheet is adaptively bent to form the required shape, which is conducive to installation. The utility model further provides a vehicle.

The application relates to the field of aquatic disease prevention and control, and discloses a Chinese soft-shelled turtle skin rot disease prevention and control method and system. The method is characterized in that a high-definition underwater camera array and a near-infrared spectrum sensing unit are deployed to collect real-time turtle body surface texture images and reflected spectrum data, and identity recognition and behavior parameter extraction are combined to input a double-flow time sequence neural network model for risk probability evaluation. When the risk value of three consecutive periods exceeds a threshold value, a slide rail type isolation hatch is automatically triggered to guide high-risk individuals to be isolated without contact, and a water quality regulation mechanism is linked to block the spread of pathogens. Through multi-modal perception, individualized tracking and intelligent decision-making closed loop, the application realizes super-early warning and precise intervention of skin rot disease, reduces the incidence rate and stress damage, improves the breeding efficiency and animal welfare, and has strong adaptability and engineering popularization value.

The invention relates to the technical field of bridge engineering and information, and discloses a steel-concrete combined bridge crack monitoring system and method based on an unmanned aerial vehicle, and the system comprises a multi-rotor unmanned aerial vehicle, a ground multi-base-station RTK-GPS system, a stereoscopic vision collection module, a laser radar scanning module and a thermal infrared imager. The multi-rotor unmanned aerial vehicle carries each monitoring module to fly to a to-be-detected area, a ground multi-base-station RTK-GPS system generates a virtual base station through carrier phase difference, and centimeter-level positioning of the unmanned aerial vehicle is achieved; the stereoscopic vision acquisition module acquires a high-definition image, the laser radar scanning module generates point cloud data, and the stereoscopic vision acquisition module and the laser radar scanning module are fused to construct a textured three-dimensional model; and the thermal infrared imager identifies hidden cracks. Millimeter-level non-contact monitoring of the crack of the joint surface of the steel web and the concrete is achieved, personnel climbing is not needed, the positioning precision is high, the adaptability is high, the method is suitable for long-term health monitoring of a bridge, and reliable data support is provided for bridge structure performance prediction.

This invention relates to a biomimetic compound eye wide-area sensing and communication system for unmanned swarms, comprising a collaborative optical signal modulation and transmission unit, a wide-area biomimetic compound eye photoelectric sensing unit, and a spatial optical communication and positioning calculation unit. By structurally encoding control commands and employing frequency shift keying modulation, digital information is converted into near-infrared modulated optical signals, enabling spatial information propagation. The receiving end utilizes a biomimetic compound eye structure to converge wide-area field-of-view optical signals and completes photoelectric conversion through a spatial position sensor. Multi-channel signal processing is combined to achieve communication signal demodulation and feature extraction. Based on this, multi-dimensional features are constructed and spatial position calculation is performed, realizing integrated information interaction and relative position perception between nodes. This invention does not rely on radio frequency communication links and satellite navigation systems. In complex electromagnetic countermeasures and navigation-constrained environments, it can achieve stable communication and collaborative sensing for unmanned swarms, possessing advantages such as strong anti-interference capability, good concealment, and high environmental adaptability.

The invention discloses a miniaturized Beidou navigation antenna based on a composite material and a preparation method of the miniaturized Beidou navigation antenna. The antenna comprises a radiation patch, a composite dielectric substrate, a metal grounding plate and a feed structure which are sequentially stacked from top to bottom, the dielectric constant epsilon r of the composite dielectric substrate is 6.1, and the dielectric loss angle tangent tan delta is 0.012; the radiation patch is arranged on the upper surface of the composite dielectric substrate, a main body of the radiation patch is of a square sheet structure, twelve cutting grooves are uniformly formed in four edges of the radiation patch, each edge comprises a first cutting groove located in the middle and second cutting grooves located in the two sides of the first cutting groove, and a corner cutting structure is further arranged in the diagonal direction of the radiation patch; the feed structure comprises a coaxial probe and an SMA adapter, and the coaxial probe penetrates from the back surface of the metal grounding plate and is connected with the radiation patch after penetrating through the composite dielectric substrate; and the radiation patch is provided with a feeding point at a position deviating from the geometric center along the y-axis direction. The antenna is compact in structure, reliable in performance and good in environmental adaptability.

The invention discloses a hyperspectral camouflage fiber fabric based on a flocking technology. The hyperspectral camouflage fiber fabric comprises an ultra-black fluff layer, an adhesive layer, a near-infrared absorption layer and an ultra-black base cloth layer which are sequentially arranged from top to bottom. Through a multi-layer synergistic structure, the fitting degree of the material and an ocean background spectrum at the wave band of 400-2500nm reaches 0.91 or above, the characteristics of weak reflection of visible light and strong absorption of near infrared light are reproduced, the problem of poor matching precision is solved, hyperspectral reconnaissance is avoided, and the concealment of ocean equipment is improved. The invention has the advantages of strong environmental adaptability, easy mass production, humidity and heat resistance of grade 1, no abnormity at-20 DEG C to 50 DEG C, and color fastness to sunlight of grade 5, and adapts to complex marine environment; based on a mature flocking technology, complex equipment is not needed, large-scale production can be achieved, and stability and economical efficiency are both considered.

The invention belongs to the field of power station and surrounding bird condition detection and defense, and particularly discloses a power station bird condition detection and defense method and system, and the system comprises a detection module, a three-defense-line module, and a monitoring center. Bird information is collected through an infrared camera, bird repelling sound waves and ultrasonic frequencies are determined according to bird species, a multi-level cooperative defense system is formed by combining a reflective mirror, ultrasonic waves, laser and various physical bird repelling devices, a monitoring center monitors in real time and dynamically adjusts bird repelling strategies, and the bird repelling effect and system adaptability are improved. According to the invention, precise, environment-friendly and sustainable power station bird situation prevention and control can be realized.

The application discloses a kind of based on cross-domain collaborative gating fusion low-illumination electric power small insulator detection method, comprising step one, constructing MFN-Net network model, step two, the visible light image and infrared image are preprocessed;Step three, obtain visible light feature by the backbone network of CSPDarknet 53 construction, step four, obtain infrared feature by cascaded HCT module, step five, visible light feature and infrared feature are fused to obtain deep semantic feature, step six, after fusion enhancement, the accurate boundary box coordinates of each detected insulator and the corresponding confidence score are output by multi-scale adaptive detection head module to deep semantic feature.This application can effectively solve the problems of small insulator feature extraction difficulty, modal interference, high miss detection and false detection rate in low-illumination scene, significantly improve the detection accuracy and robustness, provide technical support for intelligent inspection of power transmission line, with the advantages of high detection accuracy and strong environmental adaptability.

The application discloses a large-space unmanned aerial vehicle cooperative fire reconnaissance and fire extinguishing agent delivery system, and relates to the technical field of firefighting technology. The system comprises a reconnaissance unmanned aerial vehicle group, a delivery unmanned aerial vehicle group and a ground cooperative control unit. The reconnaissance aerial vehicle group synchronously collects fire field three-dimensional geometry, temperature and three-dimensional wind speed and direction information, and fuses to generate a dynamic environment parameter set. The ground control unit, based on the parameter set, predicts flow field evolution through near real-time numerical simulation, and, coupled with a fire extinguishing agent particle dynamics model, calculates an optimized unmanned aerial vehicle delivery track, hovering posture and throwing mechanism motion parameter. The delivery aerial vehicle group executes the instruction and uses a visual module to perform target accurate positioning and post-delivery track image return, forming a closed loop control, so that the accuracy, adaptability and fire extinguishing efficiency of the fire extinguishing agent delivery in a large-space complex fire field environment are significantly improved.

The application discloses a kind of based on deep learning's thermal infrared image small target detection method, this mainly first constructs high-quality true simulation dataset: adopts cyclic counter network to generate thermal infrared background image, generates simulation target in combination with point diffusion model, and synthesizes the simulation dataset with label by automatic mapping algorithm;Second, improve YOLOv8 network model, backbone network introduces windmill-shaped convolution and polarized self-attention mechanism, neck adopts bidirectional feature pyramid network, detection head adds small target layer and deletes large target layer, enhances small target feature extraction and positioning accuracy.The application solves the problems of thermal infrared small target detection dataset scarcity, target feature extraction difficulty and poor multi-scale adaptability, and improves the detection accuracy of the model on thermal infrared small target in complex background.

The invention discloses a tunnel blasting lumpiness in-situ real-time detection method based on video measurement. The method is realized through the following steps: installing cameras and high-intensity light sources on a cab, a movable arm and a bucket rod of a deslagging excavator, and matching the cameras and the high-intensity light sources; system calibration is completed by using an orientation sensor, a calibration plate and tunnel contour information; a muck pile change video is automatically collected in the deslagging process; screening typical frame segments through a change detection algorithm, and marking change types and spatio-temporal information; the disturbed rocks are restored based on the mark information, and a muck pile multi-layer three-dimensional morphological model is constructed from the outside to the inside in combination with the excavation exposure sequence; and finally extracting a lumpiness grading curve, rock spatial distribution and muck pile shape information. The system correspondingly comprises a camera, a light source, an analysis processing component, a calibration component and the like. The method solves the problems that in the prior art, the internal lumpiness of the muck pile is difficult to obtain, the detection aging is poor, and the environmental adaptability is insufficient, and automatic, instant and comprehensive detection of the tunnel blasting in-situ lumpiness is achieved.