Product warehouse missing submission system in leasing equipment

Abstract

The invention relates to the technical field of submission systems, in particular to a product warehouse missing submission system in leasing equipment, which comprises a data acquisition module, a data calculation module and an information submission module, and is characterized in that in the data calculation module, the leasing equipment is divided into areas, and m pieces of leasing equipment are arranged in each area; according to the product warehouse missing submission system in the leasing equipment, the number of missing product warehouses in the leasing equipment is combined with the people flow density of the area where the leasing equipment is located, weight metering is carried out, the comparison value is obtained, submission is carried out through the comparison value, it can be guaranteed that the priority of the area with the large missing number and the large people flow density is higher, and the missing number is large; and the priority of the area with low people flow density is reduced, so that more efficient operation of the leasing equipment is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of reporting systems, and in particular to a product warehouse missing reporting system in rental equipment. Background Art

[0002] Rental equipment, in the current market, is most common in the form of shared power banks, shared bicycles and other types of public rental shared equipment. Taking shared power banks as an example, when a power bank is rented and used, it will be shown as missing in the product warehouse. Faced with a huge amount of missing data, the current technology divides the data by region, sorts the missing information in each region where the power bank rental equipment is located, and dispatches maintenance personnel to supplement the areas with the largest number of missing items. This method is inefficient in actual use, because although some areas have a large number of missing items, the current density of people is not high, and the urgency of supplementing is not high. Although the number of missing items in some areas is slightly less than that in more areas, the density of people is greater. With a large density of people, the possibility of passing by rental equipment is higher, and the possibility of using rental equipment is also higher. Therefore, the urgency of supplementing is higher, and the existing reporting system obviously cannot report more efficiently. Summary of the invention

[0003] The purpose of the present invention is to provide a product warehouse missing reporting system in rental equipment to solve the problems raised in the above background technology.

[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: A system for reporting missing product bins in rental equipment, comprising a data acquisition module, a data calculation module and an information reporting module. In the data calculation module, the rental equipment is divided into regions, each region is provided with m rental equipment, and the number of missing product bins in each rental equipment is set to ,in , let the total amount of missing data in each region be ,but , where t is an integer greater than or equal to 1, representing different area numbers; The data acquisition module includes the number of missing product bins in each rental device Real-time numerical acquisition is performed, and the real-time numerical acquisition of the flow of people in each of the above areas is performed. Suppose the flow of people in a certain area is , let the corresponding area be , Assuming that the regional flow density is known, calculate the flow density of the area. ,but ,Right now for and The ratio of As the weight coefficient, calculate the comparison value , , where t is an integer greater than or equal to 1, indicating different area numbers. When the values ​​of t are the same, they correspond to the same area; Compare the values Arrange in order from largest to smallest, according to The corresponding regional arrangement order is obtained. Assume that the number of regions to be maintained at the same time is G. The regions ranked first G in the regional arrangement order are reported through the information reporting module, and the total amount of missing data in the corresponding region is reported at the same time. , maintenance personnel are dispatched for replenishment, and the value of G is set according to the number of maintenance personnel.

[0005] Preferably, the data acquisition module obtains numerical values ​​of the flow of people in the area through cameras, and the specific steps are as follows: the cameras are uniformly connected to the background server through network cables, and the server periodically sends time synchronization signals to all cameras in the area. After receiving the signal, the camera adjusts its own internal clock so that the image frames captured by different cameras at the same time are synchronized within the error range; the camera continuously captures RGB format video streams, divides them into independent image frames according to a predetermined frame rate, and transmits them frame by frame to the server for processing.

[0006] Preferably, the color image contains three channels of RGB, and the three-dimensional color space is mapped to the one-dimensional grayscale space through grayscale conversion to reduce the amount of calculation; then noise reduction filtering and edge enhancement processing are performed in sequence, and then human body contour detection is performed.

[0007] Preferably, in the initialization stage of the human contour detection, multiple frames of background images without human bodies are collected, and K Gaussian distribution models are established for each pixel point, which are distributed with mean, covariance and weight parameters; when a new frame arrives, the pixel value is compared with the model, and if it matches, the corresponding Gaussian distribution parameters are updated, and if it does not match, it is a foreground target. The Gaussian distribution is sorted by weight, and the weight sum greater than the set value is the background, and the rest is the foreground, and the model is dynamically updated to adapt to the gradual change of light.

[0008] Preferably, contour extraction is performed. For the foreground image, the findContours function of OpenCV is based on the contour topology traversal algorithm. Starting from the image boundary, the contour boundary pixels are tracked according to connectivity, and all closed contours are marked; the screening criteria are set, and the area threshold is set according to the minimum and maximum sizes of the human body estimated in the scene, and the perimeter threshold is used to assist in excluding irregular redundant contours; the compactness of the contour is considered through the shape factor, and those that do not conform to the aspect ratio and ellipticity of the human body are excluded, and suspected human contours are initially screened.

[0009] Preferably, the aspect ratio of the suspected human body contour is calculated, the major axis and the minor axis are determined according to the circumscribed rectangle of the contour, the circularity measures the degree of the contour's proximity to a circle, and is evaluated using the ratio of the square of the contour perimeter to the area. The data for annotating human and non-human contours are obtained through the above-mentioned feature evaluation; the annotated human and non-human contour data are used to train a support vector machine, the extracted feature vectors are input into the trained model, the discrimination result is output, the positive class is the human contour, and an accurate human contour set is constructed.

[0010] Preferably, a standard chessboard calibration object is placed on the ground of the area, the camera is used to shoot from different perspectives, the chessboard corner points are detected using OpenCV functions, and the camera intrinsic and external parameters are calculated based on the corner point coordinates using Zhang Zhengyou's calibration algorithm; after obtaining the intrinsic and external parameters, the human body contour coordinates in different camera coordinate systems are converted to a unified world coordinate system using a coordinate transformation formula.

[0011] Preferably, a cost matrix is ​​constructed, in which the rows represent the human body contours of one camera, the columns are the contours of another camera, and the elements are the quantified values ​​of the differences in position, size, and movement direction between the two. The minimum cost match is found through the Hungarian algorithm, and the same human body contours captured by different cameras are associated, and duplicate counts are eliminated. The deduplicated human body contours are summarized, and the number is the estimated value of the current number of people in the area.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The product warehouse missing reporting system for rental equipment of the present invention combines the number of missing product warehouses in the rental equipment with the pedestrian density of the area where the rental equipment is located, performs weighted measurement, and obtains a comparison value. Reporting is performed based on the comparison value, which can ensure that areas with a large number of missing products and a large pedestrian density have a higher priority, while areas with a large number of missing products and a small pedestrian density have a lower priority, thereby ensuring more efficient operation of the rental equipment.

[0013] The method in which the data acquisition module in the present invention obtains numerical values ​​of the flow of people in the area through a camera can make a more efficient and accurate estimation of the flow of people in the area, which is suitable for open areas where it is difficult to count the number of people entering and leaving, and realizes the counting of the number of people in the area. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The flowchart of the system for reporting missing product warehouses in the rental equipment of the present invention. DETAILED DESCRIPTION

[0015] 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.

[0016] See also Figure 1 The present invention provides a technical solution: a product warehouse missing reporting system in a rental device, including a data acquisition module, a data calculation module and an information reporting module. The data calculation module includes a central processing unit and a readable storage medium storing a computer program. The computer program is executed by the central processing unit. In the data calculation module, the rental equipment is divided into regions. Taking a shared power bank as an example, a number of shared power bank rental devices are located within the map range. The regions are divided on the map to ensure that the difference between the area of ​​each region and the number of rental devices is not too large. There are m rental devices in each region. Suppose the number of missing product warehouses in each rental device is ,in , let the total amount of missing data in each region be ,but , where t is an integer greater than or equal to 1, representing different area numbers; The data acquisition module includes the number of missing product warehouses in each rental device Real-time numerical acquisition is performed, as well as real-time numerical acquisition of the flow of people in each of the above-mentioned areas; the data acquisition module includes a sensor arranged at a corresponding position of the product warehouse in the rental equipment, and the sensor in the product warehouse is used to detect in real time whether the product in the product warehouse is missing. The above-mentioned sensor can adopt contact switches, photoelectric switches, and capacitive sensors, etc. When the product in the product warehouse is taken away, it can feedback a signal to detect the missing.

[0017] Suppose the traffic volume in a certain area is , let the corresponding area be , It is known that since the area divided on the map is calculated according to the map proportion, the area flow density can be calculated, and the flow density is ,but ,Right now for and The ratio of As the weight coefficient, calculate the comparison value , , where t is an integer greater than or equal to 1, indicating different area numbers. When the values ​​of t are the same, they correspond to the same area; Compare the values Arrange in order from largest to smallest, according to The corresponding regional arrangement order is obtained. Assume that the number of regions to be maintained at the same time is G. The regions ranked first G in the regional arrangement order are reported through the information reporting module, and the total amount of missing data in the corresponding region is reported at the same time. , maintenance personnel are dispatched for replenishment, the value of G is set according to the number of maintenance personnel. For example, the maintenance personnel can only be divided into 3 groups for maintenance and replenishment at the same time. At this time, the value of G is set to 3 in the system. When reporting, only the top three areas will be reported, and 3 groups of maintenance personnel will be dispatched to the corresponding areas for maintenance and replenishment; the information reporting module is composed of the Internet of Things communication system, which mainly plays a communication role.

[0018] The data acquisition module obtains the numerical value of the flow of people in the area. When the area is a closed place, such as a shopping mall, a flow counting device can be set at the entrance and exit to count the flow of people in and out. In open streets and other places, it is impossible to achieve entry and exit statistics. It is necessary to set up several cameras in the area to avoid blind spots as much as possible. The cameras are connected to the background server through network cables. The server periodically sends a time synchronization signal to all cameras in the area. After receiving the signal, the camera adjusts its own internal clock so that the image frames captured by different cameras at the same time are synchronized within the error range. The camera continuously captures RGB format video streams, divides them into independent image frames according to the established frame rate, and transmits them frame by frame to the server for processing. The transmission protocol can be HTTP, RTSP, etc.

[0019] Color images contain three channels, RGB, and the data redundancy of the three channels is high. The three-dimensional color space is mapped to the one-dimensional grayscale space through grayscale conversion to reduce the amount of calculation; then, noise reduction filtering and edge enhancement processing are performed in sequence, and then human contour detection is performed. Both noise reduction filtering and edge enhancement processing are common image processing techniques. Noise reduction filtering can use Gaussian filtering to generate a convolution kernel based on a Gaussian function. This convolution kernel is used to perform pixel-by-pixel convolution operations on the image. The central pixel value is updated to the weighted sum of the neighboring pixels. The abrupt pixel value caused by noise is smoothed, and the image quality is improved, which is conducive to subsequent precise edge extraction. Edge enhancement first smoothes the image with Gaussian filtering, and then uses Sobel operators to calculate the gradient amplitude and direction. After non-maximum suppression, only the strongest edge point in the gradient direction is retained; then through double threshold detection, those above the high threshold are retained, those below the low threshold are discarded, and the intermediate values ​​are retained depending on the neighborhood situation. Finally, the edge breakpoints are connected to obtain a coherent and clear human contour.

[0020] In the initialization stage of human contour detection, multiple frames of background images without human bodies are collected, and K Gaussian distribution models are established for each pixel. The Gaussian distribution model is a widely used probability distribution model with mean, covariance and weight parameters. When a new frame arrives, the pixel value is compared with the model. If it matches, the corresponding Gaussian distribution parameters are updated. If it does not match, it is a foreground target. The Gaussian distribution is sorted by weight. If the sum of the weights is greater than the set value, it is the background, and the rest is the foreground. The model is dynamically updated to adapt to the light gradient.

[0021] To extract contours, for foreground images, OpenCV's findContours function is based on the contour topology traversal algorithm. Starting from the image boundary, it tracks the contour boundary pixels according to connectivity and marks all closed contours. findContours is a key function in the OpenCV library for detecting the contours of objects in images. It can accurately locate the boundaries of target objects in properly preprocessed binary images. The core of the contour topology traversal algorithm based on the findContours function of OpenCV is to track the contour boundaries of target objects in images, and outline closed or non-closed contours by systematically exploring the connection relationship of pixel points. The function starts searching from the boundary pixels of the image. In the binary image obtained after image preprocessing (the target is white and the background is black, or vice versa), it moves along the boundary where the pixel value changes. For example, the exploration journey starts at the junction of a white foreground target and a black background. The pixel at this junction is the starting tracking point. The 8-connectivity rule between pixels (in a two-dimensional plane, a pixel is surrounded by 8 adjacent pixels, including the diagonal direction) or the 4-connectivity rule (only considering adjacent pixels in the horizontal and vertical directions) is used to determine which pixels belong to the same connected area, that is, the same contour.

[0022] If the pixel values ​​of adjacent pixels satisfy the conversion relationship between foreground and background and meet the connectivity settings, they are considered to belong to the same contour. Once the starting point is found, the algorithm will use recursion or iteration to continue to expand the tracking path to the surrounding pixels. Taking recursion as an example, when a boundary pixel is located, the function calls itself to check the adjacent pixels of the pixel, marks the adjacent pixels that meet the conditions as part of the contour, and continues to dig deeper until it goes back to the starting point to form a closed contour; the iterative method uses the stack structure to push the pixels to be explored into the stack, check the adjacent pixels of the pixels popped out of the stack one by one, and continuously update the stack content until the contour tracking is completed.

[0023] The screening criteria are set, and the area threshold is set according to the minimum and maximum dimensions of the human body estimated in the scene. The perimeter threshold is used to assist in excluding irregular and redundant contours. The compactness of the contour is considered through the shape factor, and those that do not conform to the length-to-width ratio and ellipticity of the human body are excluded, for the initial screening of suspected human contours.

[0024] The aspect ratio of the suspected human contour is calculated, and the major axis and minor axis are determined according to the circumscribed rectangle of the contour. The circularity measures the degree of the contour's closeness to a circle and is evaluated using the ratio of the square of the contour perimeter to the area. The data for labeling human and non-human contours is obtained through the above feature evaluation. The support vector machine is trained with the labeled human and non-human contour data, and the extracted feature vector is input into the trained model to output the discrimination result, with the positive class being the human contour, thus constructing a set of accurate human contours.

[0025] A standard chessboard calibration object is placed on the ground in the area. The camera shoots from different angles, and the OpenCV function is used to detect the corner points of the chessboard. Based on the coordinates of the corner points, the camera intrinsic and extrinsic parameters are calculated using the Zhang Zhengyou calibration algorithm. The Zhang Zhengyou calibration algorithm is a commonly used camera calibration method in this field and will not be described here. After obtaining the intrinsic and extrinsic parameters, the human body contour coordinates in different camera coordinate systems are converted to the unified world coordinate system through the coordinate transformation formula.

[0026] Construct a cost matrix, where the rows represent the human body outline of one camera, the columns represent the outline of another camera, and the elements are the quantitative values ​​of the differences in position, size, and direction of movement between the two. Use the Hungarian algorithm to find the minimum cost match, associate the same human body outline captured by different cameras, remove duplicate counts, and summarize the human body outlines after deduplication. The number is the estimated value of the current number of people in the area. The Hungarian algorithm is an existing combinatorial optimization algorithm that efficiently solves the assignment problem. It is used to best match the elements in the two sets to minimize the matching cost.

[0027] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A product warehouse missing reporting system in a rental equipment, comprising a data acquisition module, a data calculation module and an information reporting module, characterized in that: In the data calculation module, the rental equipment is divided into regions. There are m rental equipment in each region. The number of missing product warehouses in each rental equipment is set to ,in , let the total amount of missing data in each region be ,but , where t is an integer greater than or equal to 1, representing different area numbers; The data acquisition module includes the number of missing product bins in each rental device Real-time numerical acquisition is performed, and the real-time numerical acquisition of the flow of people in each of the above areas is performed. Suppose the flow of people in a certain area is , let the corresponding area be , Assuming that the regional flow density is known, calculate the flow density of the area. ,but ,Right now for and The ratio of As the weight coefficient, calculate the comparison value , , where t is an integer greater than or equal to 1, indicating different area numbers. When the values ​​of t are the same, they correspond to the same area; Compare the values Arrange in order from largest to smallest, according to The corresponding regional arrangement order is obtained. Assume that the number of regions to be maintained at the same time is G. The regions ranked first G in the regional arrangement order are reported through the information reporting module, and the total amount of missing data in the corresponding region is reported at the same time. , maintenance personnel are dispatched for replenishment, and the value of G is set according to the number of maintenance personnel.

2. A product warehouse missing reporting system in a rental equipment according to claim 1, characterized in that: The data acquisition module uses cameras to obtain numerical values ​​of pedestrian flow in the area. The specific steps are as follows: the cameras are connected to the background server through network cables. The server periodically sends time synchronization signals to all cameras in the area. After receiving the signals, the cameras adjust their internal clocks so that the image frames captured by different cameras at the same time are synchronized within the error range; the cameras continuously capture RGB format video streams, divide them into independent image frames according to the established frame rate, and transmit them frame by frame to the server for processing.

3. A product warehouse missing reporting system in a rental equipment according to claim 2, characterized in that: The color image contains three channels, RGB. The three-dimensional color space is mapped to the one-dimensional grayscale space through grayscale conversion to reduce the amount of calculation; then noise reduction filtering and edge enhancement processing are performed in sequence, and then the human body contour detection is performed.

4. A product warehouse missing reporting system in a rental equipment according to claim 3, characterized in that: In the initialization stage of the human contour detection, multiple frames of background images without human bodies are collected, and K Gaussian distribution models are established for each pixel point, which have mean, covariance and weight parameters. When a new frame arrives, the pixel value is compared with the model. If it matches, the corresponding Gaussian distribution parameters are updated. If it does not match, it is a foreground target. The Gaussian distribution is sorted by weight. If the sum of weights is greater than the set value, it is the background, and the rest is the foreground. The model is dynamically updated to adapt to the light gradient.

5. A product warehouse missing reporting system in a rental equipment according to claim 4, characterized in that: Contour extraction is performed for the foreground image. OpenCV's findContours function is based on the contour topology traversal algorithm. Starting from the image boundary, it tracks the contour boundary pixels according to connectivity and marks all closed contours. The screening criteria are set, and the area threshold is set according to the minimum and maximum size of the human body estimated in the scene. The perimeter threshold is used to assist in excluding irregular redundant contours. The compactness of the contour is considered through the shape factor, and those that do not conform to the aspect ratio and ellipticity of the human body are excluded, and suspected human contours are initially screened.

6. A product warehouse missing reporting system in a rental equipment according to claim 5, characterized in that: The aspect ratio of the suspected human contour is calculated, and the major axis and minor axis are determined according to the circumscribed rectangle of the contour. The circularity measures the degree of the contour's closeness to a circle and is evaluated using the ratio of the square of the contour perimeter to the area. The data for labeling human and non-human contours is obtained through the above feature evaluation. The support vector machine is trained with the labeled human and non-human contour data, and the extracted feature vector is input into the trained model to output the discrimination result, with the positive class being the human contour, thus constructing a set of accurate human contours.

7. A product warehouse missing reporting system in a rental equipment according to claim 6, characterized in that: Place a standard chessboard calibration object on the ground in the area, shoot with the camera from different angles, use OpenCV functions to detect the corner points of the chessboard, and calculate the camera intrinsic and extrinsic parameters based on the coordinates of the corner points using Zhang Zhengyou's calibration algorithm; After obtaining the intrinsic and extrinsic parameters, the human body contour coordinates in different camera coordinate systems are converted to the unified world coordinate system through the coordinate transformation formula.

8. A product warehouse missing reporting system in a rental equipment according to claim 7, characterized in that: Construct a cost matrix, where the rows represent the human body contours of one camera, the columns are the contours of another camera, and the elements are the quantitative values ​​of the differences in position, size, and movement direction between the two. Use the Hungarian algorithm to find the minimum cost match, associate the same human body contours captured by different cameras, eliminate duplicate counts, and summarize the deduplicated human body contours. The number is the estimated value of the current number of people in the area.

Images