Aircraft hatch cover glass surface defect detection method and equipment

Abstract

The invention provides an aircraft hatch cover glass surface defect detection method and equipment, and relates to the technical fields of machine vision, Internet of Things and edge intelligent image processing, algorithm model and state estimation and the like. The problems that manual visual detection has high requirements for experience and skills of detection personnel, labor investment is high, visual fatigue is prone to occurring, false detection and missing detection are caused, efficiency is low, and improvement of flight safety and guarantee efficiency is restricted are mainly solved. The LED light source irradiates a specific part on the surface of the glass of the hatch cover, and light rays of the LED light source are transmitted in the glass of the hatch cover and modulated by various defects including surface structure defects to form a unique image. Light rays which can not be observed by human eyes are guided to the observation surface to be imaged by the prism, wherein the light rays are totally reflected and transmitted in the hatch cover glass. The machine vision unit collects an image reflecting the surface condition of the hatch cover glass on the prism observation surface and uploads the image to the handheld terminal through the Internet of Things technology. The handheld terminal and the embedded special APP are used for detecting types of defects (such as debonding, scratches, cracks, crazing and the like) on the surface of the glass of the hatch cover, quantitatively analyzing defect characteristics and automatically judging detection data, so that a detection conclusion is formed. The method is mainly applied to detection of airplane hatch cover glass surface defects and is suitable for detection of transparent material surface defects.

Description

Technical Field

[0001] The present invention relates to the technical fields of machine vision, Internet of Things and edge intelligent image processing, algorithm model and state estimation, and in particular to a method and device for detecting defects in aircraft cockpit canopy glass. Background Art

[0002] Before the aircraft performs flight missions again, it is necessary to inspect various defects that may occur in its canopy glass, including cracks in bolt holes in the skin-covered part, debonding of structural adhesives, and debonding of polyester tapes, to ensure flight safety. The inspection of defects in aircraft canopy glass is mainly carried out through manual visual inspection and manual visual interpretation of prism imaging. There are problems such as high requirements for the experience and skills of the inspection personnel, high manual input, easy visual fatigue, false detection and missed detection, and low efficiency, which restrict the improvement of flight safety and guarantee efficiency. Summary of the invention

[0003] The purpose of the present invention is to provide a method and device for detecting surface defects of aircraft cabin glass with advanced technology, reasonable design, accurate and comprehensive detection and identification, high detection efficiency and convenient use, in view of the shortcomings of the current methods and devices for detecting surface defects of aircraft cabin glass. It mainly adopts technologies such as machine vision, Internet of Things, and edge intelligent image processing, and consists of six parts: machine vision unit, handheld terminal, dedicated APP, prism, light source and tooling.

[0004] After the present invention adopts the above technology and structure, its beneficial effects are as follows: compared with manual visual inspection, its objectivity, defect detection rate, accuracy of defect pattern recognition, and detection efficiency of aircraft cabin glass surface defects are significantly improved, and the false alarm rate is significantly reduced. Through machine vision, artificial visual fatigue is overcome and the detection accuracy is improved; the edge intelligent detection model is applied, and the attention, feature fusion, optimization loss function and model training and optimization mechanisms are introduced, so that the present invention has the ability to quickly and accurately detect defects on the surface of aircraft cabin glass, and has the ability to adapt to new types of defects; it overcomes the problems of the original detection method relying on human experience and skills, the low credibility of the detection results and the detection efficiency, and the need for regular training and assessment of the detection personnel; the comprehensive effect is to extend the maintenance and update cycle of the detection equipment, simplify the experience and skill requirements for personnel, improve the safety and attendance rate of aircraft flights, and promote the improvement of the overall benefits of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present application, but do not constitute an improper limitation of the present invention. In the drawings:

[0006] Figure 1 It is a schematic diagram of the composition of the system of the present invention;

[0007] Figure 2 and Figure 3 It is a schematic diagram of the optical path principle of the present invention.

[0008] Description of reference numerals:

[0009] 1-system, 11-machine vision unit, 111-control and communication module, 112-image acquisition module, 113-data storage module (SD card), 12-dedicated handheld terminal, 13-dedicated APP, 131-image acquisition control module, 132-image preprocessing module, 133-surface defect type recognition module, 134-surface defect quantitative analysis module, 135-main control and communication module, 14-prism, 15-LED light source, 16-tooling, 161-fixture, 162-cover; 21-hatch cover outer skin, 22-bolt hole, 23-prism, 24-hatch cover glass, 25-hatch cover glass mounting base. DETAILED DESCRIPTION

[0010] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments, wherein the illustrative embodiments and descriptions are only used to explain the present invention but are not intended to limit the present invention.

[0011] like Figure 1 As shown, a method and device for detecting surface defects of aircraft cabin cover glass described in this specific embodiment, wherein the system 1 mainly includes a machine vision unit 11, a dedicated handheld terminal 12, a dedicated APP 13, a prism 14, an LED light source 15 and a tooling 16.

[0012] The LED light source 15 is the basic light source for imaging the surface defects of the hatch cover glass 24. It irradiates a specific part of the surface of the hatch cover glass 24. Its light is transmitted inside the hatch cover glass 24 and modulated by various defects including surface structural defects to form a unique image. The function of the prism 14 is to guide the light that contains the surface defect characteristics of the hatch cover glass 24, is fully reflected and transmitted in the hatch cover glass 24, and cannot be observed by the human eye to its observation surface for imaging. The machine vision unit 11 completes the acquisition of the image reflecting the surface condition of the hatch cover glass 24 formed on the observation surface of the prism 14 due to the irradiation of the LED light source 15 on the hatch cover glass 24 under test, and uploads the image to the dedicated handheld terminal 12 through the Internet of Things technology. The dedicated handheld terminal 12 is equivalent to a small palm-sized computer with a built-in dedicated APP 13 to complete the acquisition, analysis and storage of detection data. The dedicated APP 13 has a certain number of trained and quality-assured intelligent defect detection models and quantitative analysis algorithms built in, which completes the image stitching, defect type detection (such as debonding, scratches, cracks, silver streaks, etc.) of each measurement area, quantitative analysis of defect characteristics and automatic determination of detection data to form a detection conclusion; and has a self-learning function, which can automatically repair and improve defect models and quantitative analysis algorithms. The tooling 16 is mainly used for the installation of the various components of the system 1 during the detection, and to form a local darkroom in the detection area to prevent interference from other light sources except the light emitted by the LED light source 15.

[0013] Among the constituent units of the system 1 , the prism 14 , the machine vision unit 11 , and the dedicated APP 13 are core units.

[0014] The prism 14 is precisely designed and made of a material having the same optical properties as the hatch glass.

[0015] like Figure 2 As shown, the aircraft cockpit canopy glass 24 is installed between the canopy glass mounting seat 25 and the canopy outer skin 21 through the bolt hole 22. The light incident on the canopy glass 24 (including the surrounding area near the bolt hole 22) is totally reflected by the canopy glass 24, and the human eye cannot clearly observe the bolt hole at the edge of the canopy glass and the glass state near it, the adhesive structure and the polyester tape. Whether it is intact brings great difficulties to the field inspection and maintenance.

[0016] like Figure 3 As shown, in Figure 2 On the basis of the prism 23, the light transmitted in the hatch glass 24 can be coupled to the prism 23 and emitted from the imaging surface of the prism 23, which is convenient for the machine vision unit 11 to collect images.

[0017] The machine vision unit 11 includes a control and communication module 111, an image acquisition module 112 and a data storage module 113, which complete the operation control of the unit, the acquisition of the surface defect image of the hatch cover glass 24, the data storage and the transmission of the surface defect image data of the hatch cover glass 24 to the handheld terminal 12, and has a WEB service function and supports two wireless communication modes: WIFI and BLE.

[0018] The dedicated APP 13 includes an image acquisition control module 131, an image preprocessing module 132, a surface defect type recognition module 133, a surface defect quantitative analysis module 134, and a main control and communication module 135. The image acquisition control module 131 is used to control the timing of collecting the surface defect image of the cabin cover glass 24 and the fill light intensity of the LED light source 15; the image preprocessing module 132 is used to fine-tune, align and crop the image of each measurement area; the surface defect type recognition module 133 is the HZBY-YOLO-v8 intelligent detection model; the surface defect quantitative analysis module 134 is an improved TensorFlow Lite quantitative analysis algorithm that adds a small target detection head, introduces an attention mechanism and a feature fusion mechanism, and can effectively improve the detection performance of small targets, improve computing efficiency, reduce computing resource overhead, and optimize the model real-time detection method. The main improvement methods are as follows:

[0019] First, a tiny target detection head is added to the C2f output of the Neck layer of the model structure, which can be integrated with Bi-FPN to effectively improve the detection performance of tiny defects.

[0020] Second, the triple attention mechanism is introduced to enhance the interaction of the model in obtaining channel and spatial dimension information, and improve the ability to detect tiny defects from the mixed material background; the attention mechanism is introduced after the last C2f module of the backbone, which can quickly separate defects from the background and improve computational efficiency.

[0021] Third, in response to the problems of YOLO-v8 using the FPN feature map method, the model has high requirements for computing resources and slow inference speed, the Bi-FPN mechanism, bidirectional cross-scale connection and weighted feature fusion method are introduced to enhance the feature extraction capability of the model and reduce the computing resource overhead; by adjusting the weights and fine-tuning the scale, the model's real-time detection method is optimized.

[0022] The above description is only a preferred embodiment of the present invention, so all equivalent changes or modifications made according to the structure, characteristics and principles described in the scope of the patent application of the present invention are included in the scope of the patent application of the present invention.

Claims

1. An aircraft cabin glass surface defect detection device, comprising a machine vision unit 11, a dedicated handheld terminal 12 and a dedicated APP 13, etc., characterized in that The equipment consists of a machine vision unit 11, a dedicated handheld terminal 12, a dedicated APP 13, a prism 14, an LED light source 15 and a tooling 16. The tooling 16 assembles the units to form a darkroom to inspect the test piece.

2. The aircraft cabin glass surface defect detection device according to claim 1 is characterized in that The described machine vision unit includes a control and communication module 111 , an image acquisition module 112 , and a data storage module (SD card) 113 .

3. The aircraft cabin glass surface defect detection device according to claim 1 is characterized in that The described dedicated APP 13 includes an image acquisition control module 131 , an image preprocessing module 132 , a surface defect type recognition module 133 , a surface defect quantitative analysis module 134 and a main control and communication module 135 .

4. The aircraft cabin glass surface defect detection device according to claim 1, characterized in that The image acquisition control module 131 included in the dedicated APP 13 controls the timing of acquiring the surface defect image of the hatch cover glass 24 and the fill light intensity of the LED light source 15 .

5. The aircraft cabin glass surface defect detection device according to claim 1, characterized in that The surface defect type recognition module 133 included in the dedicated APP 13 is the HZBY-YOLO-v8 intelligent detection model.

6. The aircraft cabin glass surface defect detection device according to claim 1, characterized in that The surface defect quantitative analysis module 134 included in the dedicated APP 13 is an improved TensorFlow Lite quantitative analysis algorithm that adds a small target detection head, introduces an attention mechanism and a feature fusion mechanism, and can effectively improve the detection performance of small targets, improve computing efficiency, reduce computing resource overhead, and optimize the model's real-time detection method.

Images