Search

CN-122016808-A - Unmanned aerial vehicle berth system

CN122016808ACN 122016808 ACN122016808 ACN 122016808ACN-122016808-A

Abstract

The invention discloses an unmanned aerial vehicle berthing system, which relates to the technical field of unmanned aerial vehicles and comprises an unmanned aerial vehicle body, a sensing module, a processing module and a transmission module, wherein the sensing module is a binocular vision acquisition module, the processing module is an embedded computing power processing module, a berthing area image is acquired through the binocular vision acquisition module, and the processing such as depth estimation, BEV visual angle conversion, semantic segmentation and quantitative parameter calculation is performed by combining with the embedded computing power processing module, so that microscopic defects and liquid pollutants on the surface of the berthing area can be accurately and quantitatively evaluated, the safety state of the berthing area can be definitely judged, thereby providing comprehensive and accurate surface state basis for unmanned aerial vehicle berthing, effectively avoiding potential safety hazards such as sideslip and clamping stagnation caused by unknown surface state, guaranteeing the stability and safety of the unmanned aerial vehicle berthing process, and meeting the practical application requirements of unmanned aerial vehicle high-precision safety berthing.

Inventors

  • FENG BINGWEI

Assignees

  • 元宝宝(广东)科技有限公司

Dates

Publication Date
20260512
Application Date
20260206

Claims (10)

  1. 1. The unmanned aerial vehicle berthing system comprises an unmanned aerial vehicle body, and a sensing module, a processing module and a transmission module which are arranged on the unmanned aerial vehicle body, and is characterized in that the sensing module is a binocular vision acquisition module, and the processing module is an embedded computing processing module; The unmanned aerial vehicle comprises a unmanned aerial vehicle body, a binocular vision acquisition module, an embedded computing power processing module and a transmission module, wherein the binocular vision acquisition module is used for acquiring left-eye images and right-eye images of a parking area and transmitting the left-eye images and the right-eye images to the embedded computing power processing module, the embedded computing power processing module is used for providing calibration standard, the embedded computing power processing module is used for preprocessing the left-eye images and the right-eye images, performing depth estimation on the preprocessed images to generate a depth map, performing BEV visual angle conversion on the depth map combined with the calibration standard provided by the auxiliary calibration module, performing semantic segmentation on the converted BEV images to extract surface defects and liquid pollutant characteristics, performing quantitative parameter calculation on the basis of the extracted characteristics and the depth map, performing parking safety judgment according to a quantitative parameter calculation result, and the transmission module is used for transmitting a safety judgment result to the flight control system of the unmanned aerial vehicle body.
  2. 2. The unmanned aerial vehicle docking system of claim 1, wherein when the embedded computing power processing module performs depth estimation, an improved semi-global block matching algorithm is adopted, the algorithm introduces a gradient information weight optimization cost calculation function, and sets a reflection area mask to perform neighborhood average filling on parallax values of a reflection area, and the depth estimation process further comprises dynamic parallax range adaptation processing, wherein the adaptation relation satisfies the formula: wherein For a dynamically adapted parallax range, As the reference parallax range, For the actual flight altitude of the unmanned aerial vehicle, The auxiliary calibration module comprises three calibration identification points, the embedded computing force processing module performs internal parameter calibration and external parameter calibration of the binocular vision acquisition module by acquiring images of the calibration identification points, the internal parameter comprises a focal length, main point coordinates and distortion coefficients, and the external parameter comprises a rotation matrix and a translation vector.
  3. 3. The unmanned aerial vehicle docking system of claim 1, wherein when the embedded computing power processing module performs BEV perspective conversion, an inverse projection algorithm is used to map the image pixels to the XY plane of the world coordinate system to generate the BEV image, and the mapping relationship of the inverse projection algorithm satisfies the formula: wherein And Is the coordinate of the pixel point on the world coordinate system XY plane, And Is the abscissa and ordinate of the pixel point in the image coordinate system, And As the coordinates of the principal point, Is the depth value corresponding to the pixel point, And For the focal length of the camera in the X-axis and Y-axis directions, the pixel precision of the BEV image has a fixed mapping relation with the physical size of the world coordinate system.
  4. 4. The unmanned aerial vehicle docking system of claim 1, wherein when the embedded computing power processing module performs semantic segmentation, an improved lightweight U-Net model based on MobileNetV as a main network is adopted, the model replaces a common convolution layer of an original U-Net with a depth separable convolution layer, a boundary enhancement loss function is introduced to strengthen boundary feature extraction of surface defects and liquid pollutants, an output result of the semantic segmentation comprises a background category surface defect category pit category and a liquid pollutant category, and the embedded computing power processing module is further configured to post-process the semantic segmentation result and reject isolated pixel areas with areas smaller than a set threshold.
  5. 5. The unmanned aerial vehicle docking system of claim 1, wherein the quantitative parameter calculation comprises a surface defect quantitative parameter calculation, wherein the surface defect comprises a crack and a pit, the crack quantitative parameter calculation comprises a crack width calculation, the crack width calculation comprises a step of selecting a plurality of measuring points along a crack trend in a BEV image, calculating a crack pixel width of each measuring point, converting the crack pixel width into an actual width and then taking an average value, the pit quantitative parameter calculation comprises a pit depth calculation, the pit depth calculation comprises a step of extracting a depth value of a pit area in a depth map, carrying out a difference operation with a depth value of a peripheral normal area, taking a maximum value of the difference as a pit depth, and the calculation relation satisfies the formula: wherein Is the depth of the pit slot, Is the depth value of the peripheral normal region, Is the depth value of the pit area.
  6. 6. The unmanned aerial vehicle docking system of claim 1, wherein the quantitative parameter calculation comprises a liquid pollutant quantitative parameter calculation, the liquid pollutant comprises water and oil stain, the oil stain quantitative parameter calculation comprises an oil stain coverage area calculation, the oil stain coverage area calculation is converted into an actual oil stain coverage area by counting the number of pixels of an oil stain area in a BEV image and combining a mapping relation of pixel precision and physical size, the water accumulation quantitative parameter calculation comprises a water accumulation thickness calculation, and the water accumulation thickness calculation is calculated by a refraction model by combining a depth value of the water accumulation area and a camera optical axis angle.
  7. 7. The unmanned aerial vehicle docking system according to claim 1, wherein the embedded computing power processing module is used for presetting a surface defect safety threshold and a liquid pollutant safety threshold when the docking safety judgment is carried out, comparing the calculated quantitative parameters with the corresponding safety thresholds respectively, judging that a docking area is safe to dock when all the quantitative parameters are smaller than or equal to the corresponding safety thresholds, judging that the docking area is unsafe to dock when any quantitative parameter is larger than the corresponding safety threshold, transmitting a safety judgment result and the quantitative parameters to the flight control system together by the transmission module, and calling pre-stored standby docking point coordinates when the safety judgment result and the liquid pollutant safety threshold are unsafe to dock by the embedded computing power processing module, and synchronously transmitting the standby docking point coordinates to the flight control system by the transmission module.
  8. 8. The unmanned aerial vehicle docking system of claim 1, further comprising a light supplementing module electrically connected to the embedded computing power processing module, wherein the embedded computing power processing module is configured to detect illumination intensity of an image during image preprocessing, and when the illumination intensity is lower than a set value, control the light supplementing module to turn on and adjust light supplementing brightness, wherein the image preprocessing comprises a graying process gaussian filtering process and a histogram equalization process, wherein the graying process adopts a weighted average method, and the gaussian filtering process adopts a convolution kernel with a fixed size for noise removal.
  9. 9. The unmanned aerial vehicle docking system according to claim 2 is characterized in that when the embedded computing force processing module performs internal reference and external reference calibration, a Zhang Zhengyou calibration method is adopted, a calibration equation is built to solve the internal reference and the external reference by collecting images of calibration identification points at different angles, the calibration identification points are regular patterns with alternating black and white, the embedded computing force processing module determines image coordinates of the identification points by identifying outline features of the calibration identification points, and then a mapping relation between an image coordinate system and a world coordinate system is established, and after calibration is completed, the embedded computing force processing module stores calibration parameters into a local storage unit for depth estimation and BEV visual angle conversion.
  10. 10. The unmanned aerial vehicle docking system according to claim 1, wherein the transmission module is a Wi-Fi6 transmission module, the embedded computing power processing module and the binocular vision acquisition module transmission module are used for carrying out data interaction by adopting an industrial-level communication protocol, the embedded computing power processing module is further configured to verify image data and calculation results in the transmission process, when a data transmission error is detected, a retransmission mechanism is triggered, and after the flight control system of the unmanned aerial vehicle body receives a safety judgment result, a docking path adjustment or standby docking point switching operation is carried out according to the judgment result.

Description

Unmanned aerial vehicle berth system Technical Field The invention relates to the technical field of unmanned aerial vehicles, in particular to an unmanned aerial vehicle berthing system. Background Along with the continuous development of unmanned aerial vehicle technology, the unmanned aerial vehicle is increasingly widely applied to a plurality of fields such as logistics distribution, power inspection, environment monitoring and the like, unmanned aerial vehicle autonomous berth serves as a key link in an operation flow, and the safety and reliability of the unmanned aerial vehicle directly influence the overall operation efficiency and equipment safety. In an actual application scene, the surface state of the parking area is one of the core factors for determining the parking safety, so that the surface state of the parking area is accurately evaluated, and the method is an important premise for realizing the safe and autonomous parking of the unmanned aerial vehicle. According to the unmanned aerial vehicle parking apron dynamic landing point adjusting method and system based on environment perception, the environment perception data and the flight state parameters are acquired, a three-dimensional terrain grid model is built by adopting a laser radar and vision fusion scanning mode, terrain fluctuation distribution and obstacle density distribution are analyzed, a safe landing area set is generated, and landing point positions and landing strategies are adjusted through a multipath optimization model. However, the above scheme can only realize qualitative or semi-quantitative analysis of macroscopic topography fluctuation and obstacle density, and cannot perform accurate quantitative evaluation on microscopic defects and liquid pollutants on the surface of a parking area affecting parking stability, so that comprehensive and accurate surface state basis cannot be provided for unmanned aerial vehicle parking, and actual requirements of unmanned aerial vehicle high-precision safe parking are difficult to meet. Disclosure of Invention The invention aims to make up the defects of the prior art and provides an unmanned aerial vehicle berthing system, so as to solve the problem that the prior art cannot accurately and quantitatively evaluate microscopic defects and liquid pollutants on the surface of a berthing area affecting berthing stability, and thus cannot provide comprehensive and accurate surface state basis for berthing of an unmanned aerial vehicle. The invention provides a technical scheme for solving the technical problems, which comprises an unmanned aerial vehicle berthing system, a sensing module, a processing module and a transmission module, wherein the sensing module is a binocular vision acquisition module, and the processing module is an embedded computing power processing module; The unmanned aerial vehicle comprises a unmanned aerial vehicle body, a binocular vision acquisition module, an embedded computing power processing module and a transmission module, wherein the binocular vision acquisition module is used for acquiring left-eye images and right-eye images of a parking area and transmitting the left-eye images and the right-eye images to the embedded computing power processing module, the embedded computing power processing module is used for providing calibration standard, the embedded computing power processing module is used for preprocessing the left-eye images and the right-eye images, performing depth estimation on the preprocessed images to generate a depth map, performing BEV visual angle conversion on the depth map combined with the calibration standard provided by the auxiliary calibration module, performing semantic segmentation on the converted BEV images to extract surface defects and liquid pollutant characteristics, performing quantitative parameter calculation on the basis of the extracted characteristics and the depth map, performing parking safety judgment according to a quantitative parameter calculation result, and the transmission module is used for transmitting a safety judgment result to the flight control system of the unmanned aerial vehicle body. Further, when the embedded computing power processing module executes depth estimation, an improved semi-global block matching algorithm is adopted, gradient information weight optimization cost calculation function is introduced into the algorithm, a reflective area mask is set to carry out neighborhood average filling on the parallax value of the reflective area, the depth estimation process further comprises dynamic parallax range adaptation processing, and the adaptation relation satisfies the formula: wherein For a dynamically adapted parallax range,As the reference parallax range,For the actual flight altitude of the unmanned aerial vehicle,The auxiliary calibration module comprises three calibration identification points, the embedded computing force processing module performs internal parameter calibrati