An intelligent identification device for bridge expansion joint defects

Abstract

The present invention relates to the technical field of bridge monitoring equipment, and in particular to an intelligent identification device for bridge expansion joint defects, including an identification vehicle and an identification system, wherein the identification vehicle includes a base, a shell, a shooting mechanism and a cleaning mechanism, and driving wheels and rolling wheels are respectively provided at both ends of the base so that the base can slide on the bridge, and the driving wheels are driven by a driving motor; the shell is detachably connected to the base, and the shell and the base form a placement cavity, the shooting mechanism and the cleaning mechanism are respectively located at both ends of the placement cavity, the shooting mechanism is used for shooting bridge expansion joints, and the cleaning mechanism is used for cleaning debris in bridge expansion joints; the identification system is used to obtain pictures of bridge expansion joints through the shooting mechanism to identify whether the bridge expansion joints have defects and information about the defects. The present invention can effectively identify bridge expansion joint defects and timely deal with bridge expansion defects.

Description

Technical Field

[0001] The invention relates to the technical field of bridge monitoring equipment, and in particular to intelligent identification equipment for bridge expansion joint defects. Background Art

[0002] Bridge expansion joints refer to expansion joints that are usually set between the two beam ends, between the beam ends and the abutments, or at the hinged position of the bridge to meet the requirements of bridge deck deformation. The expansion joints are required to be able to freely expand and contract in both directions parallel and perpendicular to the bridge axis, be firm and reliable, and be smooth without sudden jumps and noise when vehicles pass through; they must be able to prevent rainwater and garbage and soil from infiltrating and blocking; and they must be easy and convenient to install, inspect, maintain, and remove dirt. At the expansion joints, the railings and the bridge deck pavement must be disconnected.

[0003] Traditionally, the detection of bridge expansion joint defects is mainly completed through regular manual inspections. In practice, bridge inspection staff can first visually inspect the condition of the expansion joints, and then use a tape measure to measure if the problem is not obvious, so that they can promptly detect the accumulation of debris in the gaps of the expansion joint components, the height difference of the joints, the misalignment of the unit joints, and the abnormal sound of the gaps. At present, regular manual inspections lack real-time monitoring of indicators related to operation and maintenance safety. The management department cannot detect risks in time during the inspection intervals, and the impact of the environment cannot guarantee a good detection effect, which consumes a lot of labor costs and time costs. Summary of the invention

[0004] In order to solve the above problems, the present invention provides an intelligent identification device for bridge expansion joint defects, which can effectively identify bridge expansion joint defects and promptly deal with the bridge expansion joint defects.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] An intelligent identification device for bridge expansion joint defects includes an identification vehicle and an identification system. The identification vehicle includes a base, a shell, a shooting mechanism and a cleaning mechanism.

[0007] A driving wheel and a rolling wheel are respectively provided at both ends of the base, two driving wheels are provided, and the two driving wheels are fixedly connected by a transmission shaft, and the transmission shaft is rotatably connected to the base so that the two driving wheels are located on both sides of the base, and the transmission shaft is driven by a driving motor, and two rolling wheels are provided, and the two rolling wheels are rotatably connected to both sides of the bottom of the base away from one end of the driving wheel;

[0008] The shell is detachably connected to the base, and the shell and the base form a placement cavity. The two ends of the base are respectively provided with a first opening and a second opening connected to the placement cavity. The shooting mechanism and the cleaning mechanism are respectively located at the two ends of the placement cavity. The shooting mechanism includes a connecting frame and a camera. The connecting frame includes a first cross bar and a second cross bar. The two ends of the first cross bar are fixedly connected to the shell. The first cross bar is connected to the second cross bar through an electric telescopic rod on a side facing the base. The camera is connected to the second cross bar, and the camera faces the first opening.

[0009] The cleaning mechanism comprises a cleaning wheel and a transmission belt, wherein the cleaning wheel is located at the second opening and is rotatably connected to the housing, and two ends of the transmission belt are respectively connected to the cleaning wheel and the transmission shaft;

[0010] The recognition system includes a data acquisition module, a database and a defect recognition module. The data acquisition module is used to acquire data from pictures taken by the camera; the database is used to store data on pictures of bridge expansion joint defects and corresponding defect information; the defect recognition module is used to pair and compare the data in the data acquisition module with the data in the database to obtain whether the bridge expansion joint taken by the camera has defects and information about the defects.

[0011] Furthermore, a partition is provided in the placement cavity, and both ends of the partition are fixedly connected to the base and the outer shell respectively, and the partition divides the placement cavity into a first cavity and a second cavity, the first cavity is conductively connected to the first opening, and the shooting mechanism is located in the first cavity; the second cavity is conductively connected to the second opening, and the cleaning mechanism is located in the second cavity; a blocking brush is fixedly provided on the side of the base facing away from the placement cavity, and the blocking brush is located between the first opening and the second opening.

[0012] Furthermore, the cleaning wheel is located at one end of the forward direction of the identification vehicle, and connecting rods are fixed on both sides of the cleaning wheel, the connecting rod sleeves are provided with bearings, connecting grooves are provided on both sides of the inner wall of the second cavity, the bearings are embedded in the connecting grooves and are slidably connected to the connecting grooves, an abutment and an abutment spring are provided in the connecting groove, the abutment is slidably connected to the connecting groove, the abutment spring is located in the connecting groove, and one end of the abutment spring abuts against a side of the abutment facing away from the bearing, and the other end abuts against the connecting groove, under the action of the abutment spring, the cleaning wheel moves toward the side of the second opening facing away from the second cavity.

[0013] Furthermore, a guide cover is fixedly arranged in the second cavity, the guide cover is a funnel structure, and one end of the wide mouth of the guide cover faces the cleaning wheel; a suction mechanism is arranged on the outer wall of the shell, the suction mechanism includes a suction bucket and a suction fan blade, the suction bucket is divided into a first bin body and a second bin body by a partition net, a guide tube is arranged at one end of the first bin body away from the second bin body, one end of the guide tube is conductively connected to the first bin body, and the other end penetrates into the second cavity and is conductively connected to the narrow end of the guide cover, the suction fan blade is arranged in the second bin body, and the suction fan blade is driven by a suction motor so that the debris in the second cavity can enter the first bin body through the guide cover and the guide tube in sequence; a barrel opening connected to the first bin body is arranged on the side of the suction barrel, and the first bin body is provided with a collection box, the collection box is embedded in the first bin body through the barrel opening, and the collection box can close the barrel opening, and a mesh plate connected to the second bin body is arranged on a side of the collection box close to the partition net.

[0014] Further, the connecting frame also includes a support rod, two of the support rods are provided, and the two support rods are located on both sides of the electric telescopic rod, the support rod includes a first rod body, a second rod body and a locking mechanism, the first rod body is provided with a sliding cavity, and one end of the first rod body is provided with a sliding opening communicating with the sliding cavity, and the other end is fixedly connected to the first cross bar, one end of the second rod body is slidably embedded in the sliding cavity through the sliding opening, and the other end is fixedly connected to the second cross bar;

[0015] The locking mechanism includes a fixed shell, a locking piece and an electromagnet. The fixed shell is fixedly connected to the outer wall of the first rod body near one end of the sliding opening. The locking piece is located in the fixed shell, and one end of the locking piece is slidably connected to the inner wall of the fixed shell, and the other end penetrates into the sliding cavity. The fixed shell is provided with a locking spring, and the two ends of the locking spring are respectively abutted against the fixed shell and the locking piece, so that the locking piece can be abutted and fixed with the second rod body; an iron block is fixedly provided at one end of the locking piece away from the first rod body, and the electromagnet is arranged in the fixed shell. When the electromagnet is energized, the electromagnet can attract the iron block to move the locking piece in a direction away from the second rod body.

[0016] Furthermore, the recognition system further includes a data model building module, which is used to store a plurality of model images of bridge expansion joints with defects, and mark the defect information on each of the model images to form a data set; the data model building module is also used to train the data set, and the training steps of the data set are:

[0017] A1: Divide the model images of the data set into a training set and a test set by random sampling according to a preset ratio, and input the training set into a Laplacian convolution layer for preprocessing to obtain a preprocessed data set;

[0018] A2: constructing a target detection model for bridge expansion joint diseases through a neural network, and extracting feature layers of different scales from the preprocessed data set through the target detection model to obtain a corresponding original feature map;

[0019] A3: The target anchor point is extracted from the original feature map of step A2 as a candidate region through the anchor point frame generated by the region generation network and the softmax classifier, and the disease information of the data set is marked as a GT frame. The positive sample and the negative sample are obtained by matching the GT frame with the anchor point frame:

[0020] A4: Calculate the translation amount and scale factor between the anchor box and the GT box of the positive sample, and perform regression correction on the border of the anchor box to correct the position of the anchor box; send the corrected anchor box and the original feature map to the region of interest pooling layer, and perform maximum pooling to obtain a fixed-size proposed feature map;

[0021] A5: Obtain the probability of proposing the region of interest and the positive classification sample through the region generation network; use the proposed feature map as input to perform network training to obtain the probability vector of each type of sample and the target detection frame after the border regression in the anchor box;

[0022] A6: Repeat steps A1 to A5, set the learning rate of the shared convolutional layer in the neural network to 0, and update the unique network layer in the region generation network to retrain and obtain the detection model; input the test set into the trained detection model to determine the disease information of the bridge expansion joint image in the detection model, and store the data of the data model building module in the database.

[0023] Furthermore, in step A3, the anchor box with the largest overlap ratio with the GT box and the GT box with an overlap ratio with the candidate area greater than 0.7 are recorded as positive samples, and the GT box with an overlap ratio with any calibrated candidate area less than 0.3 is recorded as a negative sample.

[0024] Furthermore, in step A4, the proposed feature map is classified through a fully connected layer and a softmax classifier to output a probability vector of each class of samples; and the position offset of each proposal box in the neural network is obtained by using bounding box regression.

[0025] Furthermore, the housing is also provided with a GPS positioning module, and the GPS positioning module is used to obtain the location information of the identified vehicle;

[0026] The recognition system also includes a processing module, which includes a storage submodule and a suggestion submodule. The storage submodule is used to store corresponding treatment methods for bridge expansion joint defects; the suggestion submodule is used to match and compare the data of the defect recognition module with that of the storage submodule based on the position information of the recognition vehicle to obtain the disease treatment method for the bridge expansion joint at the corresponding position photographed.

[0027] Further, the first cavity is provided with a placement rack, the two ends of the placement rack are respectively fixedly connected to the inner walls of the first cavity on opposite sides, and a row of infrared sensors are evenly arranged on the surface of the placement rack facing the first opening, and the infrared sensors are used to measure the distance between the placement rack and the expansion joint of the bridge;

[0028] The recognition system also includes a flatness detection module, which is used to obtain data of the flatness detection module to calculate the height difference between the values ​​measured by adjacent infrared sensors. The flatness detection module is provided with a height difference threshold, and the flatness detection module can compare the height difference with the height difference threshold. When the height difference is greater than the height difference threshold, the flatness detection module marks the corresponding position of the picture taken by the camera.

[0029] The beneficial effects of the present invention are:

[0030] 1. By adjusting the length of the electric telescopic rod, the shooting range of the camera can be adjusted so that the pictures taken by the camera can clearly show the bridge expansion joint, so as to reduce the influence of the pictures outside the bridge expansion joint on the subsequent bridge expansion joint disease identification; at the same time, under the action of the drive motor, the identification vehicle can move at a constant speed to ensure the clarity of the pictures taken by the camera. When the drive motor is started, the transmission shaft rotates accordingly. Under the action of the transmission belt, the cleaning wheel can clean the debris in the bridge expansion joint, and at the same time, it can reduce the influence of the debris on the shooting of the bridge expansion joint. After obtaining the picture of the bridge expansion joint, the picture obtained by the camera can be matched and compared with the picture in the database and the data of the corresponding defect location and the corresponding defect type through the defect recognition module, so as to determine whether the bridge expansion joint photographed by the camera has a defect and the information of the defect, so that the monitoring personnel can timely discover the disease of the bridge expansion joint and deal with the bridge expansion disease in time, and also reduce the labor cost and time cost.

[0031] 2. Under the action of the partition and the barrier brush, the dust generated by the cleaning wheel can be prevented from affecting the shooting effect of the camera; driven by the suction motor, the suction blades generate attraction, so that the debris in the second cavity can pass through the guide cover and the guide tube into the first chamber in turn, further reducing the impact of dust on the shooting effect of the camera.

[0032] 3. With the help of the data model building module, known bridge expansion joint images and corresponding disease data can be used for training and evaluation to obtain a detection model. With the help of the detection model, the bridge expansion joint images taken by the camera can be identified through a neural network, so as to quickly determine whether there are any diseases in the bridge expansion joints and the information about the diseases.

[0033] 4. Under the action of the infrared sensor, the distance between the placement frame and the bridge expansion joint can be measured. By comparing the height difference values ​​measured by adjacent infrared sensors, it can be determined that there is a jump in the height of the two sides of the bridge expansion joint. When the height difference is greater than the height difference threshold, it proves that there is a jump in the height of the two sides of the bridge expansion joint. The leveling detection module marks the jump location, and at the same time, the disease identification module can be used to accurately determine whether the bridge expansion joint has a disease. Since the height jump on both sides of the bridge expansion joint is only a local component that cannot work normally in the early stage, the components at the bridge expansion joint where the vehicle jumps can be repaired and replaced, and the disease can be repaired in time. When the bridge expansion joint has caused the bridge deck to jump, and local repair and replacement of individual components cannot be used, the expansion joint device is replaced. Therefore, the infrared sensor and leveling detection module of the present invention can prevent the occurrence of the bridge expansion joint jump disease in time. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a structural schematic diagram of a bridge expansion joint disease intelligent identification device according to a preferred embodiment of the utility model.

[0035] Figure 2 It is a schematic diagram of the cleaning wheel structure of a bridge expansion joint disease intelligent identification device according to a preferred embodiment of the utility model.

[0036] Figure 3 It is a schematic diagram of the connection groove structure of a bridge expansion joint defect intelligent identification device according to a preferred embodiment of the utility model.

[0037] Figure 4 It is a structural schematic diagram of the photographing mechanism of a bridge expansion joint defect intelligent identification device according to a preferred embodiment of the utility model.

[0038] Figure 5 It is a schematic diagram of the connecting frame structure of a bridge expansion joint defect intelligent identification device according to a preferred embodiment of the utility model.

[0039] Figure 6 It is a schematic diagram of the structure of the attraction mechanism of the intelligent identification device for bridge expansion joint defects in a preferred embodiment of the utility model.

[0040] Figure 7It is a structural block diagram of the identification system of an intelligent identification device for bridge expansion joint defects in a preferred embodiment of the utility model.

[0041] In the figure, 1-base, 101-first opening, 102-second opening, 11-driving wheel, 111-transmission shaft, 12-rolling wheel, 13-driving motor, 14-barrier brush, 2-housing, 3-placing cavity, 301-first cavity, 302-second cavity, 303-connecting groove, 31-partition, 4-connecting frame, 401-first cross bar, 402-second cross bar, 403-electric telescopic rod, 41-camera, 42-support rod, 421-first rod body, 422-second rod body, 423-sliding cavity, 43-fixed shell, 431-locking piece, 432-electromagnet, 433-locking Spring, 434-iron block, 5-cleaning wheel, 51-transmission belt, 52-guide cover, 53-suction barrel, 530-partition net, 531-first warehouse body, 532-second warehouse body, 533-barrel mouth, 54-suction fan blade, 541-suction motor, 55-guide tube, 56-collection box, 561-screen plate, 61-data acquisition module, 62-database, 63-disease identification module, 64-data model building module, 65-processing module, 651-storage submodule, 652-suggestion submodule, 66-GPS positioning module, 67-leveling detection module, 7-placing rack, 71-infrared sensor. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0043] It should be noted that when a component is referred to as being "fixed to" another component, it may be directly on the other component or there may also be a component centered. When a component is considered to be "connected to" another component, it may be directly connected to the other component or there may also be a component centered. When a component is considered to be "set on" another component, it may be directly set on the other component or there may also be a component centered. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0045] Please also see Figures 1 to 7 A preferred embodiment of the present invention provides an intelligent identification device for bridge expansion joint defects, including an identification vehicle and an identification system.

[0046] The identification vehicle includes a base 1, a housing 2, a shooting mechanism and a cleaning mechanism.

[0047] A driving wheel 11 and a rolling wheel 12 are respectively provided at both ends of the base 1. Two driving wheels 11 are provided, and the two driving wheels 11 are fixedly connected by a transmission shaft 111. The transmission shaft 111 is rotatably connected to the base 1 so that the two driving wheels 11 are located on both sides of the base 1. The transmission shaft 111 is driven by a driving motor 13. Two rolling wheels 12 are provided, and the two rolling wheels 12 are rotatably connected to both sides of the bottom of the base 1 away from the driving wheel 11. The rolling wheel 12 of this embodiment is a universal wheel to facilitate the adjustment of the forward angle of the identification vehicle. The driving motor 13 is fixed to the base 1.

[0048] The housing 2 is detachably connected to the base 1, and the housing 2 and the base 1 form a placement cavity. The two ends of the base 1 are respectively provided with a first opening 101 and a second opening 102 connected to the placement cavity. The shooting mechanism and the cleaning mechanism are respectively located at the two ends of the placement cavity. The shooting mechanism includes a connecting frame 4 and a camera 41. The connecting frame 4 includes a first crossbar 401 and a second crossbar 402. The two ends of the first crossbar 401 are fixedly connected to the housing 2. The first crossbar 401 is connected to the second crossbar 402 through an electric telescopic rod 403 on the side facing the base 1. The camera 41 is connected to the second crossbar 402, and the camera 41 faces the first opening 101. The second crossbar 402 of this embodiment is also provided with an illumination lamp 411.

[0049] By adjusting the length of the electric telescopic rod 403, the shooting range of the camera 41 can be adjusted so that the picture taken by the camera 41 can clearly show the bridge expansion joint, thereby reducing the impact of the picture outside the bridge expansion joint on the subsequent identification of bridge expansion joint defects; at the same time, under the action of the drive motor 13, the identification vehicle can move at a uniform speed to ensure the clarity of the picture taken by the camera 41.

[0050] The connecting frame 4 also includes a support rod 42, two support rods 42 are provided, and the two support rods 42 are located on both sides of the electric telescopic rod 403, the support rod 42 includes a first rod body 421, a second rod body 422 and a locking mechanism, the first rod body 421 is provided with a sliding cavity 423, and one end of the first rod body 421 is provided with a sliding opening connected to the sliding cavity 423, and the other end is fixedly connected to the first cross bar 401, one end of the second rod body 422 is slidably embedded in the sliding cavity 423 through the sliding opening, and the other end is fixedly connected to the second cross bar 402.

[0051] The locking mechanism includes a fixed shell 43, a locking piece 431 and an electromagnet 432. The fixed shell 43 is fixedly connected to the outer wall of the first rod body 421 near the sliding opening. The locking piece 431 is located in the fixed shell 43, and one end of the locking piece 431 is slidably connected to the inner wall of the fixed shell 43, and the other end penetrates into the sliding cavity 423. The fixed shell 43 is provided with a locking spring 433, and the two ends of the locking spring 433 are respectively abutted against the fixed shell 43 and the locking piece 431, so that the locking piece 431 can be abutted and fixed with the second rod body 422; an iron block 434 is fixedly provided at one end of the locking piece 431 away from the first rod body 421, and the electromagnet 432 is arranged in the fixed shell 43. When the electromagnet 432 is energized, the electromagnet 432 can attract the iron block 434 to move the locking piece 431 in a direction away from the second rod body 422.

[0052] The electromagnet 432 of this embodiment is electrically connected to the electric telescopic rod 403 through the controller. When the electric telescopic rod 403 is started, the controller electromagnet 432 is energized, so that the electromagnet 432 can attract the iron block 434, and the locking piece 431 releases the second rod body 422 to allow the electric telescopic rod 403 to freely extend and retract. When the electric telescopic rod 403 stops, the controller electromagnet 432 loses power, and under the action of the locking spring 433, the locking piece 431 moves toward the second rod body 422 and abuts and fixes against the second rod body 422, so as to improve the stability of the electric telescopic rod 403 when it stops, and ensure that the pictures taken by the camera 41 are clear.

[0053] The housing of this embodiment is further provided with a push rod, on which a touch screen for controlling the driving motor 13, the camera 41 and the electric telescopic rod 403 is provided.

[0054] The cleaning mechanism includes a cleaning wheel 5 and a transmission belt 51 . The cleaning wheel 5 is located at the second opening 102 and is rotatably connected to the housing 2 . Two ends of the transmission belt 51 are respectively connected to the cleaning wheel 5 and the transmission shaft 111 .

[0055] When the driving motor is started, the transmission shaft 111 rotates accordingly, and under the action of the transmission belt 51, the cleaning wheel 5 can clean the debris in the bridge expansion joint, and at the same time can reduce the impact of the debris on the shooting of the bridge expansion joint.

[0056] In this embodiment, a partition 31 is further provided in the placement cavity, and the two ends of the partition 31 are fixedly connected to the base 1 and the housing 2 respectively, and the partition 31 divides the placement cavity into a first cavity 301 and a second cavity 302. The first cavity 301 is conductively connected to the first opening 101, and the shooting mechanism is located in the first cavity 301; the second cavity 302 is conductively connected to the second opening 102, and the cleaning mechanism is located in the second cavity 302; a blocking brush 14 is fixedly provided on the side of the base 1 facing away from the placement cavity, and the blocking brush 14 is located between the first opening 101 and the second opening 102. Under the action of the partition 31 and the blocking brush 14, dust generated by the cleaning wheel can be prevented from affecting the shooting effect of the camera 41.

[0057] The cleaning wheel 5 is located at one end of the identification vehicle in the forward direction, and connecting rods 501 are fixedly provided on both sides of the cleaning wheel 5, and bearings 502 are sleeved on the connecting rods 501. Connecting grooves 303 are provided on both sides of the inner wall of the second cavity 302, and the bearings 502 are embedded in the connecting grooves 303 and are slidably connected with the connecting grooves 303. Abutment members 503 and abutment springs 504 are provided in the connecting grooves 303, and the abutment members 503 are slidably connected with the connecting grooves 303. The abutment springs 504 are located in the connecting grooves 303, and one end of the abutment springs 504 abuts against the abutment member 503 facing away from the bearing 502, and the other end abuts against the connecting grooves 303. Under the action of the abutment springs 504, the cleaning wheel 5 moves toward the second opening 102 facing away from the second cavity 302. Because the cleaning wheel 5 has a buffering property, the cleaning wheel 5 can adapt to expansion joints of different depths, and under the action of the abutment springs 504, the cleaning ability of the cleaning wheel 5 can be improved.

[0058] In this embodiment, a guide cover 52 is fixedly arranged in the second cavity 302, and the guide cover 52 is a funnel structure, and the wide end of the guide cover 52 faces the cleaning wheel 5; the outer wall of the shell 2 is provided with a suction mechanism, and the suction mechanism includes a suction bucket 53 and a suction fan blade 54, and the suction bucket 53 is divided into a first bin body 531 and a second bin body 532 by a partition net 530, and a guide pipe 55 is provided at one end of the first bin body 531 away from the second bin body 532, and one end of the guide pipe 55 is conductively connected to the first bin body 531, and the other end penetrates into the second cavity 302 and is conductively connected to the narrow end of the guide cover 52, so as to attract The fan blades 54 are arranged in the second warehouse body 532, and the suction fan blades 54 are driven by the suction motor 541, so that the debris in the second cavity 302 can enter the first warehouse body 531 through the guide cover 52 and the guide tube 55 in sequence; the side of the suction barrel 53 is provided with a barrel mouth 533 connected to the first warehouse body 531, and the first warehouse body 531 is provided with a collection box 56, the collection box 56 is embedded in the first warehouse body 531 through the barrel mouth 533, and the collection box 56 can close the barrel mouth 533, and the collection box 56 is provided with a mesh plate 561 connected to the second warehouse body 532 on the side close to the partition net 530.

[0059] When the identification vehicle moves forward, the cleaning wheel 5 is located in front of the camera 41, so that the camera 41 can take pictures of the expansion joint of the bridge after cleaning. Since the cleaning wheel 5 will raise dust when rotating, in this embodiment, under the drive of the suction motor 541, the suction blade 54 generates suction, so that the debris in the second cavity 302 can pass through the guide cover 52 and the guide tube 55 in turn into the first warehouse body 531, further reducing the influence of dust on the shooting effect of the camera 41.

[0060] In this embodiment, a storage battery 15 is provided on the base 1 to supply power to electrical appliances on the identification vehicle.

[0061] The identification system includes a data acquisition module 61, a database 62, a disease identification module 63 and a data model building module 64.

[0062] The data acquisition module 61 is used to acquire data of the pictures taken by the camera 41. The data acquisition module 61 of this embodiment acquires the data of the camera 41 by wireless transmission.

[0063] The database 62 is used to store the data of the bridge expansion joint defect images and the corresponding defect information. The defect information of this embodiment includes the location of the bridge where the defect expansion joint is located, the defect location of the bridge expansion joint, and the type of the defect.

[0064] The defect recognition module 63 is used to compare the data of the data acquisition module 61 with the data of the database 62 to obtain whether the bridge expansion joint photographed by the camera 41 has defects and the information of the defects.

[0065] After obtaining the picture of the bridge expansion joint, the picture obtained by the camera can be matched and compared with the picture in the database 62 and the data of the corresponding defect location and the corresponding defect type through the defect recognition module 63, so as to determine whether there is a defect in the bridge expansion joint photographed by the camera and the information of the defect, so that the monitoring personnel can discover the defect of the bridge expansion joint in time and deal with the defect of the bridge expansion joint in time, and also reduce labor cost and time cost.

[0066] The data model building module 64 is used to store multiple model images of bridge expansion joints with defects, and mark the defect information for each model image to form a data set. The defect information in this embodiment includes the location of the bridge where the defective expansion joint is located, the defect location of the bridge expansion joint, and the type of defect.

[0067] The data model building module 64 is also used for training the data set. The training steps of the data set are:

[0068] A1: Divide the model images of the data set into a training set and a test set by random sampling according to a preset ratio, input the training set into the Laplace convolution layer for preprocessing to obtain a preprocessed data set. In this embodiment, the ratio of the training set to the test set is 4:1.

[0069] A2: A target detection model for bridge expansion joint defects is constructed through a neural network, and feature layers of different scales are extracted from the preprocessed data set through the target detection model to obtain the corresponding original feature map.

[0070] A3: The anchor box generated by the region generation network and the softmax classifier extract the target anchor point from the original feature map of step A2 as the candidate region, and mark the disease information of the dataset as the GT box. By matching the GT box with the anchor box, positive samples and negative samples are obtained.

[0071] In step A3, the anchor box with the largest overlap ratio with the GT box and the GT box with an overlap ratio greater than 0.7 with the candidate region are recorded as positive samples, and the GT box with an overlap ratio with any calibration in the candidate region less than 0.3 is recorded as a negative sample.

[0072] A4: Calculate the translation and scale factor between the anchor box and the GT box of the positive sample, and perform regression correction on the border of the anchor box to correct the position of the anchor box; send the corrected anchor box and the original feature map to the region of interest pooling layer, and perform maximum pooling to obtain a fixed-size proposal feature map.

[0073] In step A4, the proposed feature map is classified through a fully connected layer and a softmax classifier to output a probability vector for each class of samples; and the position offset of each proposal box in the neural network is obtained by using bounding box regression to regress a more accurate target detection box.

[0074] A5: Obtain the probability of proposing regions of interest and positively classified samples through the region generation network; use the proposed feature map as input for network training to obtain the probability vector of each type of sample and the target detection box after the bounding box in the anchor box is regressed.

[0075] A6: Repeat steps A1 to A5, set the learning rate of the shared convolutional layer in the neural network to 0, and update the unique network layer in the regional generation network to retrain and obtain the detection model; input the test set into the trained detection model to determine the disease information of the bridge expansion joint image in the detection model, and store the data of the data model building module 64 in the database 62.

[0076] In this embodiment, the housing 2 is further provided with a GPS positioning module 66 , which is used to obtain the position information of the identified vehicle; and the GPS positioning module 66 can mark the position information on the picture taken by the camera 41 .

[0077] The identification system further includes a processing module 65, which includes a storage submodule 651 and a suggestion submodule 652. The storage submodule 651 is used to store corresponding treatment methods for bridge expansion joint diseases.

[0078] The suggestion submodule 652 is used to match and compare the data of the defect recognition module 63 with the data of the storage submodule 651 according to the position information of the recognition vehicle, so as to obtain a method for treating the defects of the bridge expansion joint at the corresponding position photographed.

[0079] Since the damage treatment methods for expansion joints at different positions on the bridge are different, this embodiment obtains the geographic information of the pictures taken by the camera 41 through the GPS positioning module 66, so that the suggestion submodule 652 can obtain the location of the bridge where the defective expansion joint is located and the corresponding damage information, and accurately obtain the damage treatment method of the bridge expansion joint based on the data of the storage submodule 651, ensuring the correct treatment of the damage.

[0080] In this embodiment, the first cavity 301 is provided with a placement rack 7, and the two ends of the placement rack 7 are respectively fixedly connected to the inner walls of the first cavity 301 on opposite sides, and a row of infrared sensors 71 are evenly arranged on the surface of the placement rack 7 facing the first opening 101, and the infrared sensors 71 are used to measure the distance between the placement rack 7 and the expansion joint of the bridge;

[0081] The recognition system also includes a flatness detection module 67, which is used to obtain data of the flatness detection module 67 to calculate the height difference between the values ​​measured by adjacent infrared sensors 7. The flatness detection module 67 is provided with a height difference threshold, and the flatness detection module 67 can compare the height difference with the height difference threshold. When the height difference is greater than the height difference threshold, the flatness detection module 67 marks the corresponding position of the picture taken by the camera 41.

[0082] Under the action of the infrared sensor 71, the distance between the placement frame 7 and the bridge expansion joint can be measured. By comparing the height difference values ​​measured by adjacent infrared sensors 7, it can be determined that there is a jump in the height of the two sides of the bridge expansion joint. When the height difference value is greater than the height difference threshold, it proves that there is a jump in the height of the two sides of the bridge expansion joint. The leveling detection module 67 marks the jump location, and at the same time, the disease identification module 63 can be used to accurately determine whether there is a disease in the bridge expansion joint. Since the height jump on both sides of the bridge expansion joint is only a local component that cannot work normally in the early stage, the components at the bridge expansion joint where the vehicle jumps can be repaired and replaced, and the disease can be repaired in time. When the bridge expansion joint has caused the bridge deck to jump, and local repair and replacement of individual components cannot be used, the expansion joint device is replaced. Therefore, the infrared sensor 71 and the leveling detection module 67 of the embodiment can prevent the occurrence of the bridge expansion joint jump disease in time.

[0083] When using the intelligent identification device for bridge expansion joint defects of this embodiment, the identification vehicle is placed on the bridge expansion joint, the length of the electric telescopic rod 403 is controlled so that the camera 41 obtains a suitable field of view, the drive motor 13 is controlled to start, the identification vehicle moves forward, and the cleaning wheel 5 cleans the bridge expansion joint. When the identification vehicle completes walking on the bridge expansion joint, a picture of the bridge expansion joint is obtained.

[0084] The data acquisition module 61 acquires the picture taken by the camera 41, and the defect recognition module 63 compares the data acquisition module 61 with the data in the database 62 to obtain whether the bridge expansion joint photographed by the camera 41 has defects and information about the defects, so that the defect recognition module 63 can remind when there are defects in the bridge expansion joint.

[0085] When the defect identification module 63 determines that there is a defect in the expansion joint of the bridge, the suggestion submodule 652 obtains the geographic information corresponding to the bridge expansion joint image through the GPS positioning module 66. The suggestion submodule 652 searches for a suitable defect treatment method in the storage submodule 651 according to the location of the bridge where the expansion joint defect is located, the location of the expansion joint where the defect is located, and the type of defect, thereby ensuring the correct treatment of the defect.

[0086] The flatness detection module 67 marks the corresponding position of the picture obtained by the camera 41, and the defect recognition module 63 indicates the position where the height jump exists on both sides of the corresponding bridge expansion joint.

Claims

1. An intelligent identification device for bridge expansion joint defects, characterized in that: It comprises an identification vehicle and an identification system, wherein the identification vehicle comprises a base (1), a housing (2), a photographing mechanism and a cleaning mechanism. A driving wheel (11) and a rolling wheel (12) are respectively provided at both ends of the base (1); two driving wheels (11) are provided, and the two driving wheels (11) are fixedly connected via a transmission shaft (111); the transmission shaft (111) is rotationally connected to the base (1), so that the two driving wheels (11) are located on both sides of the base (1); the transmission shaft (111) is driven by a driving motor (13); two rolling wheels (12) are provided, and the two rolling wheels (12) are rotationally connected to both sides of the bottom of the base (1) away from the driving wheel (11); The housing (2) is detachably connected to the base (1), and the housing (2) and the base (1) form a placement cavity. The base (1) is provided with a first opening (101) and a second opening (102) at two ends thereof, respectively, which are communicated with the placement cavity. The shooting mechanism and the cleaning mechanism are respectively located at the two ends of the placement cavity. The shooting mechanism comprises a connecting frame (4) and a camera (41). The connecting frame (4) comprises a first cross bar (401) and a second cross bar (402). The two ends of the first cross bar (401) are fixedly connected to the housing (2). A side of the first cross bar (401) facing the base (1) is connected to the second cross bar (402) via an electric telescopic rod (403). The camera (41) is connected to the second cross bar (402), and the camera (41) faces the first opening (101). The cleaning mechanism comprises a cleaning wheel (5) and a transmission belt (51); the cleaning wheel (5) is located at the second opening (102), and the cleaning wheel (5) is rotatably connected to the housing (2); and two ends of the transmission belt (51) are respectively transmission-connected to the cleaning wheel (5) and the transmission shaft (111); The identification system comprises a data acquisition module (61), a database (62) and a defect identification module (63); the data acquisition module (61) is used to acquire data from a picture taken by the camera (41); the database (62) is used to store data of a picture of a bridge expansion joint defect and corresponding defect information; the defect identification module (63) is used to compare the data from the data acquisition module (61) with the data from the database (62) to obtain whether the bridge expansion joint photographed by the camera (41) has a defect and the defect information; The recognition system further comprises a data model building module (64), wherein the data model building module (64) is used to store a plurality of model images of bridge expansion joints with defects, and to mark the defect information on each of the model images to form a data set; the data model building module (64) is also used to train the data set, and the training steps of the data set are: A1: Divide the model images of the data set into a training set and a test set by random sampling according to a preset ratio, and input the training set into a Laplacian convolution layer for preprocessing to obtain a preprocessed data set; A2: constructing a target detection model for bridge expansion joint diseases through a neural network, and extracting feature layers of different scales from the preprocessed data set through the target detection model to obtain a corresponding original feature map; A3: The target anchor point is extracted from the original feature map of step A2 as a candidate region through the anchor point frame generated by the region generation network and the softmax classifier, and the disease information of the data set is marked as a GT frame. The positive sample and the negative sample are obtained by matching the GT frame with the anchor point frame: A4: Calculate the translation amount and scale factor between the anchor box and the GT box of the positive sample, and perform regression correction on the border of the anchor box to correct the position of the anchor box; send the corrected anchor box and the original feature map to the region of interest pooling layer, and perform maximum pooling to obtain a fixed-size proposed feature map; A5: Obtain the probability of proposing regions of interest and positively classified samples through the region generation network; The proposed feature map is used as input for network training to obtain a probability vector of each type of sample and a target detection frame after the bounding box in the anchor box is regressed; A6: Repeating steps A1 to A5, setting the learning rate of the shared convolutional layer in the neural network to 0, and updating the unique network layer in the region generation network to retrain and obtain the detection model; inputting the test set into the trained detection model to determine the disease information of the bridge expansion joint image in the detection model, and storing the data of the data model building module (64) in the database (62); A partition (31) is also provided in the placement cavity, and two ends of the partition (31) are fixedly connected to the base (1) and the shell (2) respectively, and the partition (31) divides the placement cavity into a first cavity (301) and a second cavity (302), the first cavity (301) is conductively connected to the first opening (101), and the shooting mechanism is located in the first cavity (301); the second cavity (302) is conductively connected to the second opening (102), and the cleaning mechanism is located in the second cavity (302); a blocking brush (14) is fixedly provided on a side of the base (1) facing away from the placement cavity, and the blocking brush (14) is located between the first opening (101) and the second opening (102); The housing (2) is also provided with a GPS positioning module (66), and the GPS positioning module (66) is used to obtain the position information of the identified vehicle; The recognition system further comprises a processing module (65), wherein the processing module (65) comprises a storage submodule (651) and a suggestion submodule (652), wherein the storage submodule (651) is used to store corresponding treatment methods for bridge expansion joint defects; and the suggestion submodule (652) is used to match and compare the data of the defect recognition module (63) with the data of the storage submodule (651) according to the position information of the recognition vehicle, so as to obtain the treatment method for the defect of the bridge expansion joint at the corresponding position photographed; The first cavity (301) is provided with a placement rack (7), the two ends of the placement rack (7) are respectively fixedly connected to the inner walls of the first cavity (301) on two opposite sides, and a row of infrared sensors (71) are evenly arranged on the surface of the placement rack (7) facing the first opening (101), and the infrared sensors (71) are used to measure the distance between the placement rack (7) and the bridge expansion joint; The recognition system further comprises a flatness detection module (67), wherein the flatness detection module (67) is used to obtain data of the flatness detection module (67) to calculate a height difference between values ​​measured by adjacent infrared sensors (71); the flatness detection module (67) is provided with a height difference threshold, and the flatness detection module (67) is capable of comparing the height difference with the height difference threshold; when the height difference is greater than the height difference threshold, the flatness detection module (67) marks a corresponding position of a picture obtained by the camera (41) in response to the shooting.

2. The intelligent identification device for bridge expansion joint defects according to claim 1 is characterized by: The cleaning wheel (5) is located at one end of the identification vehicle in the forward direction, and connecting rods (501) are fixedly provided on both sides of the cleaning wheel (5), and the connecting rods (501) are sleeved with bearings (502). Connecting grooves (303) are provided on both sides of the inner wall of the second cavity (302), and the bearings (502) are embedded in the connecting grooves (303) and are slidably connected to the connecting grooves (303). An abutment member (503) and an abutment spring (504) are provided in the connecting grooves (303). The abutment member (503) is slidably connected to the connecting groove (303), the abutment spring (504) is located in the connecting groove (303), and one end of the abutment spring (504) abuts against a side of the abutment member (503) facing away from the bearing (502), and the other end abuts against the connecting groove (303). Under the action of the abutment spring (504), the cleaning wheel (5) moves toward the second opening (102) and the side facing away from the second cavity (302).

3. The intelligent identification device for bridge expansion joint defects according to claim 1 is characterized by: A guide cover (52) is fixedly arranged in the second cavity (302), the guide cover (52) being a funnel structure, and a wide end of the guide cover (52) faces the cleaning wheel (5); a suction mechanism is arranged on the outer wall of the housing (2), the suction mechanism comprising a suction bucket (53) and a suction fan blade (54); the suction bucket (53) is divided into a first bin body (531) and a second bin body (532) by a partition net (530); a guide tube (55) is arranged at one end of the first bin body (531) away from the second bin body (532); one end of the guide tube (55) is conductively connected to the first bin body (531), and the other end penetrates into the second cavity (302) and is conductively connected to the narrow end of the guide cover (52); the suction fan blade (54) is arranged at The suction fan blade (54) is inside the second bin body (532), and the suction fan blade (54) is driven by the suction motor (541), so that the debris in the second cavity (302) can enter the first bin body (531) through the guide cover (52) and the guide tube (55) in sequence; the side of the suction barrel (53) is provided with a barrel opening (533) connected to the first bin body (531), and the first bin body (531) is provided with a collection box (56), the collection box (56) is embedded in the first bin body (531) through the barrel opening (533), and the collection box (56) can close the barrel opening (533), and the collection box (56) is provided with a mesh plate (561) connected to the second bin body (532) on a side of the collection box (56) close to the partition net (530).

4. The intelligent identification device for bridge expansion joint defects according to claim 1 is characterized by: The connecting frame (4) further comprises a support rod (42), wherein two support rods (42) are provided and the two support rods (42) are located on both sides of the electric telescopic rod (403), wherein the support rod (42) comprises a first rod body (421), a second rod body (422) and a locking mechanism, wherein the first rod body (421) is provided with a sliding cavity (423), and one end of the first rod body (421) is provided with a sliding opening communicating with the sliding cavity (423), and the other end is fixedly connected to the first cross bar (401), and one end of the second rod body (422) is slidably embedded in the sliding cavity (423) through the sliding opening, and the other end is fixedly connected to the second cross bar (402); The locking mechanism comprises a fixed shell (43), a locking member (431) and an electromagnet (432); the fixed shell (43) is fixedly connected to the outer wall of the first rod body (421) near one end of the sliding opening; the locking member (431) is located in the fixed shell (43); one end of the locking member (431) is slidably connected to the inner wall of the fixed shell (43); the other end of the locking member (431) penetrates into the sliding cavity (423); the fixed shell (43) is provided with a locking spring (433); both ends of the locking spring (433) are respectively connected to the fixed shell (43); The fixed shell (43) and the locking piece (431) are in abutment with each other, so that the locking piece (431) can be fixedly abutted with the second rod body (422); an iron block (434) is fixedly provided at one end of the locking piece (431) away from the first rod body (421); the electromagnet (432) is arranged in the fixed shell (43); when the electromagnet (432) is energized, the electromagnet (432) can attract the iron block (434) to move the locking piece (431) in a direction away from the second rod body (422).

5. The intelligent identification device for bridge expansion joint defects according to claim 1 is characterized by: In step A3, the anchor box with the largest overlap ratio with the GT box and the GT box with an overlap ratio with the candidate area greater than 0.7 are recorded as positive samples, and the GT box with an overlap ratio with any calibration in the candidate area less than 0.3 is recorded as a negative sample.

6. The intelligent identification device for bridge expansion joint defects according to claim 1 is characterized by: In step A4, the proposed feature map is classified through a fully connected layer and a softmax classifier to output a probability vector of each class of samples; and the position offset of each proposal box in the neural network is obtained by using bounding box regression.

Images