A parking gate pole state detection system and method

Abstract

The present invention discloses a detection system and method for the state of a parking barrier pole, including the following technologies and methods: during the use of the barrier pole, a camera with more than 2 million pixels is installed in front and behind the barrier pole (or oblique) in front (or behind), and the angle and position are adjusted to fully cover the barrier pole in the video area; the image capture, image preprocessing algorithm, recognition algorithm model, and barrier pole judgment method software are deployed to an edge computing terminal (located in a parking lot) or a cloud service; during use, an image is captured every second by an image capture algorithm, the area is calculated by a recognition algorithm model, and the barrier pole state is calculated by a barrier pole judgment method, and an early warning is issued when it is normally open, deformed, or damaged. In the above manner, the present invention can help unmanned parking lots to promptly discover the operating status and damage of barrier poles, improve the operating efficiency of parking lots, and reduce manual patrols and fee evasion.

Description

Technical Field

[0001] The present invention relates to the field of intelligent parking lots, and in particular to a detection system and method for the state of a barrier pole. Background Art

[0002] Unmanned parking facilitates daily life and improves traffic efficiency. In the ever-developing modern society, the technology is becoming more mature. Unmanned operation is the standard we pursue in technology, but some potential problems still need us to solve in depth to achieve true unmanned operation.

[0003] The license plate system has a high level of unmanned intelligence, but we cannot perceive the operating status of the barrier, which is a device that is easily damaged by car collisions in actual scenarios. Generally, manual inspection is used, but it is time-consuming, labor-intensive and inefficient. There is a method to detect the barrier status by installing sensors on the barrier, but it is difficult to adapt to different barrier locations and has high installation requirements, and can only sense the open and closed states. There is also a method [1] that uses image features for template matching, but there are several disadvantages that make it very practical: 1. There is only one frame for the open and closed template images, which leads to large changes in features under different lighting conditions during the day and night, rain, fog and sunny days, resulting in template matching failure; 2. The precise pixel coordinates of the barrier pole cannot be accurately located, so a large number of features containing background interference are extracted for comparison; 3. Only the open and closed states are included and the fracture and deformation states cannot be judged. Summary of the invention

[0004] The main technical problem solved by the present invention is to provide a detection system and method for the status of a barrier pole, which can solve the problem of toll evasion caused by the failure of the barrier pole to close in time due to damage to the barrier pole or gate failure in an unattended parking lot.

[0005] In order to solve the above technical problems, a technical solution adopted by the present invention is: to provide a detection system and method for the status of a parking barrier pole, including the following technologies and methods: when the barrier pole is in use, cameras with more than 2 million pixels are installed in front and behind the barrier pole (or obliquely) in front (or behind), and the angle and position are adjusted to completely cover the barrier pole in the video area; the image capture, image preprocessing algorithm, recognition algorithm model, and barrier pole judgment method software are deployed to an edge computing terminal (located in the parking lot) or a cloud service; during use, an image is captured every second by the image capture algorithm, the area is calculated by the recognition algorithm model, and then the barrier pole status is calculated by the barrier pole judgment method, and an early warning is issued when it is normally open, deformed, or damaged.

[0006] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0007] The existing technology mainly relies on manual polling of the cloud video monitoring mode to check the switch of the barrier pole and its deformation and damage. The present invention can automatically realize the early warning of the operating status and damage status of the barrier pole of the unmanned parking lot through artificial intelligence algorithm, improve the accuracy of confirming the normal status of the barrier pole, reduce the online manual monitoring cost and offline manual patrol cost and personnel fatigue, improve the equipment operation efficiency, and reduce the evasion of fees. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a three-dimensional structural schematic diagram of a preferred embodiment of a detection system and method for the state of a barrier pole of the present invention;

[0009] Figure 2 The present invention is a technical schematic diagram of a system flow decomposition of a detection system and method for the state of a barrier pole;

[0010] Figure 3 A schematic diagram of a semantic segmentation algorithm network structure shown in a detection system and method for the state of a barrier pole;

[0011] Figure 4 The present invention is a schematic diagram of a state judgment method based on ROI area calculation, shown in a detection system and method for the state of a barrier pole. DETAILED DESCRIPTION

[0012] The preferred embodiments of the present invention are described in detail below in conjunction with the accompanying drawings so that the advantages and features of the present invention can be more easily understood by those skilled in the art, thereby making a clearer and more definite definition of the protection scope of the present invention.

[0013] The present invention is implemented including:

[0014] 1. The present invention provides a camera and barrier pole installation and deployment structure based on image recognition, wherein cameras with more than 2 million pixels are installed in front of (or obliquely in front of) the barrier pole, and the angle and position are adjusted so that the barrier pole can be completely covered in the video area. Figure 1 The three-dimensional structural schematic diagram of a preferred embodiment is shown.

[0015] 2. The present invention provides a system solution for barrier pole identification. The processing flow is as follows: Figure 2As shown in the technical diagram of the system process decomposition, the state of the barrier pole is accurately identified through the deep learning image recognition algorithm to reduce the cost of online manual monitoring and offline manual patrol costs and personnel fatigue, improve equipment operation efficiency, and reduce fee evasion. In terms of the specific process, the codec module is used to extract the image frame of the video stream, and then the filtering and image enhancement are used to improve the image quality. The rain and fog removal algorithm is combined to optimize the recognition effect. Then, the image semantic segmentation algorithm is used to accurately identify the two states of the barrier pole, the open and closed states, and the pixel point set of the barrier pole in the image is extracted. Then the coordinates of the circumscribed polygon of all the pixel points of the barrier pole are obtained to calculate the area of ​​the barrier pole. Finally, it is compared with the pre-collected calibrated area. If the current area is smaller than the calibrated area, it is judged to be damaged and deformed. If it is judged to be an abnormal deformation, the staff will be notified to perform maintenance.

[0016] 3. The present invention provides a semantic segmentation network structure such as Figure 3 As shown in the figure, the ResNeXt-101 residual network is used as the feature extraction network, and the FPN feature pyramid is used to mine multi-scale information. The combination of ResNeXt-101 and FPN makes the feature mining expression more powerful. The backbone convolutional network Resnet-C4 maps the region of interest ROI generated by the region proposal network RPN to the output of Resnet-C4, and performs ROI Poling region of interest pooling operation on it. Finally, the pooled results are parallel forked to predict the "class" category, "box" bounding box, and "mask" segmentation mask respectively.

[0017] 4. The present invention provides a method for determining the state of a gate pole based on ROI area calculation. Figure 4 As shown in the figure, firstly, the input barrier pole video stream frame image is processed and converted into the RGB format required by the network, then the full convolution semantic segmentation network model is initialized to complete the memory space allocation and network parameter loading operations, and then the full-size segmentation model inference is completed to obtain the lifting or closing state of the barrier pole, and at the same time obtain the specific pixel point set of the ROI area of ​​the barrier pole that does not contain the background, and then the pixel point set is fitted with the external contour to obtain the polygon coordinates and calculate the ROI polygon area. We have already calibrated and calculated the calibration area of ​​the barrier pole in the normal state (open, closed) in advance. Since there is no deformation in the normal state, the size of the barrier pole pixel point set and the external contour area in the open and closed states are relatively consistent. Therefore, by comparing the current ROI area with the calibration area, we can know whether it is deformed. However, in the actual processing process, the barrier pole may be interfered by mud, dust layer, etc., and the prediction of a small number of pixels is not so accurate. Therefore, we introduce threshold judgment, that is, when the absolute value of the difference between the current ROI area and the calibration area is greater than the threshold, it is judged as deformation, and if it is less than the threshold, it is normal.

[0018] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for detecting the state of a parking barrier pole, characterized in that: A. Installation method: Install a camera with more than 2 million pixels in front of, diagonally in front of, behind or diagonally behind the barrier pole, and adjust the angle and position so that the barrier pole can be fully covered in the video area; B. Image capture: The image frames in the video stream are captured by the video stream decoding module for subsequent image barrier pole recognition processing; C. Image preprocessing: In order to improve the recognition effect, a filtering algorithm is used to suppress noise blur, and image enhancement and rain and fog removal algorithms are used to solve the impact of extreme weather conditions on recognition. D. Recognition algorithm model: intercept the barrier pole image through the video stream, perform pixel-level semantic segmentation training through the fully convolutional deep network, train a suitable barrier pole image segmentation algorithm model, and calculate several states of the barrier pole and the geometric area of ​​the barrier pole; the several states include raised and closed; E. Barrier pole state judgment method: By comparing with the area storing the correct barrier pole state, the barrier pole state is output, including lifting, closing and deformation; F. During the use of the barrier pole: deploy the image capture, image preprocessing, algorithm recognition, and barrier pole status judgment method software to the edge computing terminal or cloud service. During use, the image is captured every second through the image capture algorithm, the area is calculated through the recognition algorithm model, and then the barrier pole status is calculated through the barrier pole status judgment method. In case of normal opening, deformation, or damage, an early warning is issued; The semantic segmentation algorithm accurately identifies the open and closed states of the barrier pole, extracts the pixel set of the barrier pole in the image, and then calculates the coordinates of the circumscribed polygon of all the barrier pole pixels to calculate the area of ​​the barrier pole. The semantic segmentation algorithm uses the ResNeXt-101 residual network as the feature extraction network, the FPN feature pyramid mines multi-scale information, and the backbone convolutional network Resnet-C4 maps the region of interest ROI generated by the region proposal network RPN to the output of Resnet-C4, and performs ROI Poling region of interest pooling operation on it, and finally parallel forks the pooled results; The barrier pole state judgment method is based on ROI area calculation. First, the input barrier pole video stream frame image is processed and converted into the RGB format required by the network. Then, the full convolution semantic segmentation network model is initialized to complete the memory space allocation and network parameter loading operations. Then, the full-size segmentation model is completed to obtain the lifting or closing state of the barrier pole. At the same time, the specific pixel point set of the ROI area of ​​the barrier pole that does not contain the background is obtained. Next, the pixel point set is fitted with an external contour to obtain the polygon coordinates, and the ROI polygon area is calculated. The current ROI area is compared with the calibrated area to determine whether the barrier pole is deformed.

2. A method for detecting the state of a parking barrier pole according to claim 1, characterized in that: The image capture method performs image enhancement under complex weather conditions and illumination environments to increase image clarity, and uses rain and fog removal algorithms to solve the problem of extreme climate scene recognition.

3. A method for detecting the state of a parking barrier pole according to claim 1, characterized in that: The recognition algorithm model, after learning through the fully convolutional deep network, can solve the different image qualities in severe weather conditions and complex environments with different illumination during the day and at night. The image segmentation algorithm model is optimized to obtain the accurate area.

Images