CN-122024348-A - Automatic acquisition method for inspection images of distribution network unmanned aerial vehicle of lightweight AI

CN122024348ACN 122024348 ACN122024348 ACN 122024348ACN-122024348-A

Abstract

The invention discloses an automatic acquisition method of an inspection image of a distribution network unmanned aerial vehicle based on a lightweight AI, which comprises the following steps of 1, starting running of the unmanned aerial vehicle, obtaining the inspection image when reaching an inspection shooting point, 2, positioning target power grid equipment in the inspection image based on a target detection and identification algorithm, outputting characteristic parameters of the target power grid equipment, 3, calculating position deviation of the target equipment and the center of the inspection image according to the characteristic parameters of the target power grid equipment, adjusting the posture of a cradle head and the position of the unmanned aerial vehicle according to the position deviation to enable the target equipment to be located in the center of the inspection image, 4, obtaining illumination characteristics through an ambient light sensor, adjusting a camera parameter combination through an automatic dimming algorithm based on the ambient illumination characteristics, and then executing photographing action to obtain the image of the target power grid equipment. According to the invention, the acquisition of the images of the key components of the network distribution inspection can be completed without manual intervention, and the one-time shooting qualification rate and the operation efficiency of the inspection images are obviously improved.

Inventors

LI ZHEDONG
LIANG HAO
SHU YI
AN YUNZHAN
WANG GUOFENG
MA TAIYI
YU JIAMIAO

Assignees

国网浙江省电力有限公司绍兴供电公司

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (8)

1. The automatic acquisition method for the inspection images of the distribution network unmanned aerial vehicle based on the lightweight AI is characterized by comprising the following steps: step 1, starting operation of an unmanned aerial vehicle, and obtaining a patrol image when the unmanned aerial vehicle reaches a patrol shooting point; step 2, positioning target power grid equipment in the inspection image based on a target detection and identification algorithm, and outputting characteristic parameters of the target power grid equipment; Step 3, calculating the position deviation of the target equipment and the center of the inspection image according to the characteristic parameters of the target power grid equipment, and adjusting the posture of the cradle head and the position of the unmanned aerial vehicle according to the position deviation so that the target equipment is positioned in the center of the inspection image; Step 4, acquiring illumination characteristics through an ambient light sensor, adjusting a camera parameter combination through an automatic dimming algorithm based on the ambient illumination characteristics, and then executing a photographing action to acquire a target power grid equipment image; and 5, the unmanned aerial vehicle reaches the next inspection shooting point, and the steps 1 to 4 are repeatedly executed until the target power grid equipment images of all the inspection shooting points are acquired, so that the automatic acquisition process of the inspection images of the unmanned aerial vehicle in the distribution network is completed.
2. The automatic acquisition method of the inspection image of the distribution network unmanned aerial vehicle based on the lightweight AI according to claim 1, wherein the step 2 is characterized in that the target power grid equipment is positioned in the inspection image based on a target detection and identification algorithm, and the characteristic parameters of the target power grid equipment are output, specifically: the target detection and recognition algorithm is YOLOv or YOLOv algorithm, the characteristic information of the inspection image is detected and matched with the preset power grid equipment characteristics, and the position coordinate, the size duty ratio and the attitude angle data of the target power grid equipment are output.
3. The automatic acquisition method of the inspection image of the distribution network unmanned aerial vehicle based on the lightweight AI according to claim 2, wherein the step 3 is characterized in that according to the characteristic parameters of the target power grid equipment, the position deviation of the target equipment and the center of the inspection image is calculated, and according to the position deviation, the posture of the cradle head and the position of the unmanned aerial vehicle are adjusted to enable the target equipment to be positioned in the center of the inspection image, and specifically comprises the following steps: The method comprises the steps of continuously tracking target power grid equipment in a continuous inspection image sequence based on characteristic parameters of the target power grid equipment, comparing position coordinates of the target power grid equipment with central coordinates of the inspection images, calculating position deviations of the target power grid equipment in the horizontal direction and the vertical direction, judging a horizontal offset state of the target power grid equipment by combining an attitude angle of the target power grid equipment, driving a cradle head to rotate by using a control algorithm, and adjusting the relative position of an unmanned aerial vehicle, so that the target power grid equipment is positioned in the center of the inspection images and keeps the horizontal attitude, and the position deviations and the horizontal offsets do not exceed preset thresholds.
4. The automatic acquisition method of the inspection image of the distribution network unmanned aerial vehicle based on the lightweight AI of claim 3, wherein the control algorithm is an event-triggered circulation control algorithm based on a holder posture position error threshold, the error threshold comprises a horizontal direction position deviation smaller than 5 pixels, a vertical direction position deviation smaller than 5 pixels and a horizontal offset angle smaller than 0.5, and when the deviation or the offset exceeds the error threshold, the control algorithm triggers the adjustment action of the holder and the unmanned aerial vehicle.
5. The automatic acquisition method of the inspection image of the distribution network unmanned aerial vehicle based on the lightweight AI according to claim 1, wherein in the step 4, an ambient light sensor is arranged on the unmanned aerial vehicle, illumination characteristics comprise a brightness average value, a contrast ratio, a high-light pixel duty ratio, a dark pixel duty ratio and a backlight or forward light scene mark of a target power grid equipment area, the illumination characteristics are input into a preset automatic dimming algorithm, the automatic dimming algorithm judges the illumination type of the current unmanned aerial vehicle scene through multi-dimensional evaluation, and an adaptive camera parameter combination is dynamically output according to the illumination type.
6. The automatic acquisition method of inspection images of a distribution network unmanned aerial vehicle based on lightweight AI according to claim 5, wherein the illumination type comprises normal, backlight or strong light.
7. The automatic acquisition method of the inspection images of the unmanned aerial vehicle with the distribution network based on the lightweight AI of claim 5, wherein the camera parameter combination comprises a camera ISO, a shutter speed and exposure compensation, the ISO is controlled within a low noise threshold range, the shutter speed is matched with the flight stability requirement of the unmanned aerial vehicle, and the exposure compensation is finely adjusted to a target area.
8. The automatic acquisition method of the inspection image of the distribution network unmanned aerial vehicle based on the lightweight AI according to claim 1, wherein in the step 4, before the shooting action is executed to acquire the image of the target power grid device, the camera is automatically zoomed, specifically: calculating the blurring degree of the inspection image through Laplacian, judging that the image is blurred if the blurring degree exceeds a set threshold value, refocusing, restarting the camera to refocus if the blurring degree still exceeds the set threshold value after refocusing, judging that the lens has faults if the blurring degree still exceeds the set threshold value, and reminding related personnel.

Description

Automatic acquisition method for inspection images of distribution network unmanned aerial vehicle of lightweight AI Technical Field The invention relates to the technical field of intelligent inspection of power distribution networks, in particular to an automatic acquisition method for inspection images of unmanned aerial vehicles of a distribution network with lightweight AI. Background With the continuous expansion of the distribution network scale, the distribution network line inspection is used as a key link for guaranteeing the power supply reliability, and the operation intensity and the quality requirements are increasingly improved. The conventional distribution network inspection mainly relies on manual pole climbing operation or ground visual inspection, so that the efficiency is low, safety risks such as high-altitude falling and electric shock are faced in complex terrain environments such as mountain areas, lake and river coasts, meanwhile, the burden of basic-level inspection personnel is heavy, and the operation and maintenance requirements of large-scale distribution networks are difficult to adapt. In order to solve the pain point of traditional manual inspection, unmanned aerial vehicle inspection technology is gradually popularized and applied in distribution network inspection, but the prior art scheme still has obvious defects, so that inspection image quality is unstable, data value is not fully released, and the method is specifically expressed as follows: The existing unmanned aerial vehicle inspection target alignment link is required to be completed manually by a flying hand, and the flying hand is required to simultaneously consider the unmanned aerial vehicle flight control and the aiming work of the key parts of the pole tower. The method is influenced by factors such as operation experience of a flying hand, working fatigue, hand shake and the like, and is easy to cause that a target component deviates from the center of an image, the gesture is inclined, and even a key inspection part is omitted, so that subsequent image auditing cannot be carried out because target positioning is not up to standard. The network inspection scene is complex, and the environment is often faced with various illumination conditions such as backlight, strong light, cloudy days and the like, and the camera parameter adjustment of the existing unmanned aerial vehicle mainly depends on the manual operation of a flight. The fly is difficult to determine the environment illumination change in real time and accurately and match with the optimal parameters, so that the collected image is easy to have the problems of high light overexposure or dead black of a dark part, the detail characteristics of key parts are lost, and the requirements of defect identification or manual image inspection cannot be met. The existing unmanned aerial vehicle inspection mostly adopts fixed focus shooting or flying manual zooming modes, and long-range view coverage and short-range view details are difficult to be considered. The manual zooming operation is complicated, focusing blurring is easily caused by data transmission delay and operation errors of a fly hand, and an automatic judging and correcting mechanism for image blurring is absent, so that once focusing failure occurs, the fly and the supplementary shooting are required to be carried out again, and the inspection efficiency is greatly reduced. The links of target identification, alignment, parameter adjustment, focusing, photographing and the like of the existing unmanned aerial vehicle inspection are mutually independent, and manual intervention and coordination are needed, so that a 'man-pull shoulder' type operation mode is formed. The inspection process is fragmented, the inspection time of a single-base tower is long, the image one-time shooting qualification rate is low, first-line personnel need to frequently and repeatedly supplement and fly, the data utilization rate is low, the core requirement of basic level burden reduction cannot be effectively met, and the digital and intelligent transformation process of distribution network inspection is restricted. Disclosure of Invention The invention aims to overcome the defects of inaccurate target positioning, unbalanced illumination adaptation, fuzzy focusing and the like of images caused by manual intervention of target positioning, parameter adjustment and focusing due to lack of an automatic collaborative image acquisition mechanism in the conventional distribution network unmanned aerial vehicle inspection technology, and provides a light-weight AI distribution network unmanned aerial vehicle inspection image automatic acquisition method. The invention aims at realizing the following technical scheme: A method for automatically collecting inspection images of a distribution network unmanned aerial vehicle based on lightweight AI comprises the following steps: step 1, starting operation of an unmanned