Hospital crowd density estimation method and system
By combining the hospital information system to obtain real data, determine the population density level and select appropriate estimation methods, the problem of large errors in population density estimation in the prior art is solved, and the accuracy and early warning efficiency of density estimation are improved.
Patent Information
- Application Number
- PCT/CN2023/095203
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-08
- Filing Date
- 2023-05-19
- Publication Date
- 2024-11-14
AI Technical Summary
The prior art has large errors in estimating the population density of hospitals, making it difficult to accurately estimate in different density scenarios, resulting in low efficiency of early warning systems.
By obtaining real data in combination with the hospital information system, determining the population density level, and selecting an estimation method based on pixel features or texture features based on the level, improving the accuracy of population density estimation.
The accuracy of population density estimation and the efficiency of early warning system are improved, ensuring accuracy in different density scenarios.
Smart Images

Figure CN2023095203_14112024_PF_FP_ABST
Abstract
Description
A hospital crowd density estimation method and system Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for estimating hospital crowd density. Background Art
[0002] At present, hospitals, especially large tertiary hospitals, are usually overcrowded during peak hours. If the crowd density is too high and not controlled, stampede accidents are very likely to occur. To prevent risks from occurring, it is particularly important to estimate the crowd density in hospitals through video surveillance systems and establish early warning monitoring systems.
[0003] Common crowd density estimation methods fall into two categories: pixel-based and texture-based. Pixel-based methods are the most primitive. Their fundamental principle is that there is a strong positive correlation between the number of crowd members and the number of pixels. More people correspond to more pixels. Therefore, the number of crowd members can be estimated by counting the number of pixels in a crowd target. However, this method is only applicable to scenes with low crowd density. Texture-based methods use texture features as density features. Their fundamental principle is that crowd images of different densities possess different texture patterns, and crowd density can be classified based on these texture patterns. However, at low crowd densities, texture features are less discriminative, leading to larger classification errors. Therefore, this method is primarily applicable to scenes with higher crowd density. Therefore, selecting an effective crowd density estimation method to achieve fast and efficient crowd density estimation has become one of the many research challenges in practical early warning and monitoring systems.
[0004] Summary of the Invention
[0005] In response to the above problems, the present invention provides a hospital crowd density estimation method and system to solve the technical problem of inaccurate crowd density assessment in existing medical systems.
[0006] The technical solutions provided by the present invention are as follows:
[0007] In one aspect, the present invention provides a method for estimating hospital crowd density, comprising:
[0008] Determine the area of estimated population density;
[0009] Obtaining the number of people and the area of the crowd density estimation area;
[0010] estimating the crowd density in the crowd density estimation area based on the number of people and the area;
[0011] determining a density level of the crowd density estimation area based on a crowd density estimate;
[0012] Determining a crowd density estimation method according to the number of people and density level in the crowd density estimation area, wherein the crowd density estimation is a crowd density estimation method based on pixel features or a crowd density estimation method based on texture features;
[0013] Acquiring image information of the crowd density estimation area;
[0014] The acquired image information is processed according to a determined crowd density estimation method, and the crowd density in the crowd density estimation area is estimated.
[0015] In another aspect, the present invention provides a hospital crowd density estimation system, comprising:
[0016] a camera device, disposed in a crowd density estimation area, and capturing image information of the crowd density estimation area;
[0017] Hospital Information Systems;
[0018] A crowd density assessment terminal is connected to the shooting device and the hospital information system respectively, and is used to estimate the crowd density area according to the determination; obtain the number of people and the area of the crowd density estimation area based on the hospital information system; estimate the crowd density of the crowd density estimation area based on the number of people and the area; determine its density level based on the crowd density estimation of the crowd density estimation area; determine a crowd density estimation method according to the number of people and the density level of the crowd density estimation area, wherein the crowd density estimation is a crowd density estimation method based on pixel features or a crowd density estimation method based on texture features; and process the acquired image information according to the determined crowd density estimation method to estimate the crowd density of the crowd density estimation area.
[0019] The hospital crowd density estimation method and system provided by the present invention can at least bring about the following beneficial effects: combining the real data obtained in real time by the hospital information system, establishing a correspondence between the crowd image and the real data of the hospital, preliminarily determining the density level of the crowd density estimation area, and then selecting a more efficient crowd density estimation method based on the determined density level, thereby improving the accuracy of crowd density estimation and improving the warning efficiency of the early warning system. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] The preferred implementation scheme will be described below in a clear and understandable manner with reference to the accompanying drawings to further illustrate the above characteristics, technical features, advantages and their implementation methods.
[0021] FIG1 is a flow chart of a method for estimating hospital population density according to the present invention;
[0022] FIG2 is a structural diagram of a hospital population density estimation system according to the present invention.
[0023] Reference numerals: 10 - imaging equipment, 20 - hospital information system, 30 - crowd density assessment terminal. DETAILED DESCRIPTION
[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific embodiments of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings and other embodiments can be obtained based on these drawings without inventive work.
[0025] The first embodiment of the present invention is a hospital crowd density estimation method, as shown in Figure 1, including: S10 determining a crowd density estimation area; S20 obtaining the number of people and the area of the crowd density estimation area; S30 estimating the crowd density in the crowd density estimation area based on the number of people and the area; S40 determining the density level of the crowd density estimation area based on the estimated crowd density; S50 determining a crowd density estimation method based on the number of people and the density level in the crowd density estimation area, the crowd density estimation being a crowd density estimation method based on pixel features or a crowd density estimation method based on texture features; S60 obtaining image information of the crowd density estimation area; S70 processing the obtained image information according to the determined crowd density estimation method, and estimating the crowd density in the crowd density estimation area.
[0026] With the rapid development of hospital information technology, hospital appointment registration systems, outpatient check-in systems, examination appointment systems, imaging or ultrasound triage systems, and ward bed management systems have become widespread. Hospitals have also generally established integrated service platforms, enabling data interconnection across these systems. These information systems often contain real-time data from areas prone to congestion, such as hospital outpatient waiting areas, examination waiting areas, and payment waiting areas. Based on this, this embodiment establishes a correspondence between surveillance images of the crowd density estimation area and real-world data from the hospital information system. This allows for a preliminary determination of the density level of the crowd density estimation area, and then selects a more efficient crowd density estimation method based on the determined density level, thereby improving the accuracy of crowd density estimation.
[0027] The crowd density estimation area can be a specific area within the hospital, such as the outpatient waiting area, the examination waiting area, the payment waiting area, and other areas prone to crowd congestion, or it can be an outpatient building and other areas. In the application, it can be determined based on the actual situation of the hospital.
[0028] The number of people in the crowd density estimation area can be the number of people in the crowd density estimation area obtained in real time, or it can be the average number of people in the crowd density estimation area at multiple moments within a pre-set period of time (such as 1 minute, 5 minutes, etc.). In the application, the method that best reflects the actual situation can be selected to determine the number of people in the crowd density estimation area based on the actual situation.
[0029] The area of the region is the area of the crowd density estimation area. This area can be determined based on the hospital's architectural conditions or pre-delineated and stored based on actual conditions. For example, a certain area at the payment window can be designated as the payment waiting area. Before estimating the crowd density, each crowd density estimation area and its corresponding area are determined and stored. This allows the corresponding area to be searched and read after a crowd density estimation area is selected during the crowd density estimation process.
[0030] In estimating the crowd density R, two methods such as formula (1) and (2) can be used according to the situation:
[0031] Where People is the number of people in the crowd density estimation area, m is the area of the area, and λ is the preset coefficient;
[0032] Among them, People1 is the number of people entering the crowd density estimation area, and People2 is the number of people leaving the crowd density estimation area.
[0033] Regarding the selection of the above two methods, if the hospital information system can directly determine the number of people in the corresponding crowd density estimation area, such as the outpatient waiting area, the examination waiting area, etc., the number of people in the corresponding area can be predicted based on the number of people signed in in the system, and the crowd density R can be directly calculated using formula (1). If the number of people in the corresponding crowd density estimation area cannot be directly determined based on the hospital information system, or the determination process is relatively complicated, such as the outpatient building, the number of people inside can be directly determined based on the number of people entering and leaving the building (including patients and accompanying personnel, etc.), that is, the crowd density R can be determined using formula (2). Due to the influence of factors such as headcount and area area, for example, some areas of the building area such as the waiting area are inaccessible to the crowd, the above two methods also introduce a preset coefficient λ to regulate the calculation results. The preset coefficient λ needs to be determined according to the actual situation and is usually set between 0 and 1. In practical applications, since the area of the crowd density estimation region changes in real time, when estimating the crowd density, it is often necessary to estimate the crowd density for a certain period of time, such as 1 minute, 5 minutes, etc., to obtain the crowd density range of the period, so as to provide a more accurate basis for subsequent estimation.
[0034] In determining the density level based on the crowd density estimation in the crowd density estimation area, when the estimated crowd density is 0.8 people / m 2 If the number of people per 13 square meters is less than 10, the crowd density in the estimated area is low; when the estimated crowd density is higher than 0.8 people / m 2 Or if the number of people per 13 square meters is greater than 10, the crowd density in the estimated area is too high. See Table 1 for details:
[0035] Table 1 Density level classification standards
[0036] Based on the above density level classification standards, when the crowd density in the crowd density estimation area is judged to be low, the crowd density estimation method based on pixel features is selected for estimation; when the crowd density in the crowd density estimation area is judged to be high, the crowd density estimation method based on texture features is selected for estimation.
[0037] When the crowd density estimation is a crowd density estimation method based on pixel features, the acquired image information is processed according to the determined crowd density estimation method, and the crowd density of the crowd density estimation area is estimated, including: S71 extracting the foreground image of the image information; S72 calculating the foreground pixel area of the foreground image; S73 performing edge detection on the foreground image and counting the number of edge pixels; S74 establishing linear models between the number of people in the crowd density estimation area and the foreground pixel area and the number of edge pixels based on the least squares method; S75 estimating the crowd density of the crowd density estimation area based on the linear module.
[0038] Extracting effective crowd features is the key to estimating crowd density, and the premise of crowd feature extraction is to obtain effective crowd targets. In this embodiment, the preprocessing of video images mainly includes noise removal, background subtraction, foreground image, and morphological processing, and finally obtains crowd targets.
[0039] (1) Background extraction
[0040] The average background modeling method is used in background extraction. The foreground target is regarded as image noise. The grayscale values of the corresponding pixels of a certain number of frames in a continuous video are accumulated and averaged to smooth the foreground information and obtain the background image of the scene. The specific algorithm process is as follows:
[0041] 1. Extract N consecutive frames of image f from the video image sequence i (x, y), i = 1, 2, ... N;
[0042] 2. The grayscale values of the corresponding pixels (x, y) of the sampled N frames are accumulated and averaged, and the background B(x, y) image is:
[0043] The more images are sampled, the more times the pixel belongs to the background, and the calculated pixel average is closer to the background.
[0044] (2) Background subtraction
[0045] Perform differential processing on the current frame image and the clean background image that has been extracted. The part of the current frame that is close to the background will be removed, while the grayscale value of the foreground area will be retained because it is significantly different from the background grayscale value, thus obtaining a differential image containing the foreground target. Selecting an appropriate threshold to binarize the differential image can obtain the crowd foreground image. The differential image binarization formula is as follows:
[0046] Where f(x,y) represents the pixel value at (x,y) in the difference image, T is the binarization threshold, and F(x,y) represents the binary image. By setting an appropriate threshold T, the difference image f(x,y) is divided into background and foreground regions based on the relationship between the pixel value and the threshold.
[0047] The pixel area characteristics and edge pixel characteristics of the foreground object will change with the number of people in the image. When the crowd density increases, the area and edge length of the foreground image will increase accordingly, and vice versa. The present invention uses two pixel characteristics, the pixel area characteristics and edge pixel characteristics of the foreground object, to characterize the foreground image, and uses the least squares method to establish a linear model between the number of people and these two characteristics, and uses this linear model to estimate the crowd density of the crowd image. Specifically:
[0048] (1) Foreground pixel area calculation
[0049] The pixel area feature of the foreground target refers to the area of the pixels occupied by the foreground target area. In order to facilitate calculation and model building, the area feature can be replaced by the area ratio of the pixels of the foreground target area to the entire image, which ranges from 0% to 100%. When calculating, you only need to count the number of pixels of the foreground target and divide it by the number of pixels of the entire image to obtain the pixel area ratio.
[0050] (2) Calculation of the number of edge pixels
[0051] The edge pixel feature of the foreground object mainly refers to the number of edge pixels of the foreground object. It is usually detected using a differential operator. Since the Canny operator extracts the best edge effect and has high edge positioning accuracy, this embodiment uses the Canny operator to calculate the number of edge pixels. The Canny operator first uses a suitable Gaussian filter to remove noise in the image, and then uses non-maximum suppression technology to process the image to obtain edge information. Its algorithm mainly consists of four steps:
[0052] ① Use Gaussian filter to smooth the image, that is, perform convolution operation on the image and Gaussian smoothing filter, as shown in formula (4): g(x,y)=h(x,y,σ)*f(x,y) (4)
[0053] Where g(x,y) represents the smoothed image; * represents convolution; and h(x,y,σ) is a Gaussian filter: σ represents the standard deviation of the image.
[0054] ② Use the first-order differential to calculate the intensity and direction of the gradient, and the calculation formula is as shown in equations (5) and (6): θ(x,y)=arctan[Q(x,y) / P(x,y)] (6)
[0055] Among them, P(x,y) and Q(x,y) represent the first-order differences of the point (x,y) in the x-direction and y-direction respectively, M(x,y) represents the strength of the gradient, and θ(x,y) represents the direction of the gradient.
[0056] ③ Use non-maximum suppression technology to process the obtained gradient.
[0057] The non-maximum suppression technique is to compare the gradient value M(x,y) at the point (x,y) with the gradient values of the pixels around it. If the gradient value of the point is not a local maximum, the gradient value M(x,y) of the point is set to 0, otherwise no change is made.
[0058] ④ Use the dual-threshold algorithm to detect and connect edges, and select two thresholds (high threshold and low threshold) with the help of the hysteresis rule. It is considered that all pixels with grayscale greater than the high threshold are definitely edge pixels, and pixels with grayscale greater than the low threshold depend on whether they are combined with pixels greater than the high threshold. If they are adjacent, they are also considered edge pixels.
[0059] (3) Least squares curve fitting
[0060] The foreground pixel area ratio feature of the image can be obtained by counting the number of pixels of the extracted crowd foreground targets and then dividing it by the total number of pixels of the entire image. The foreground edge pixel number feature of the image can be obtained by counting the pixels of the extracted foreground edge. When performing curve fitting, 40 images are first extracted from the surveillance video as sample images for curve fitting. Then, the crowd foreground targets of the sample images are extracted, and the foreground pixel area ratio and the number of foreground edge pixels of each sample image are counted, which are marked as x1 respectively. i and x2 i Then, we use manual methods to count the actual number of people in each sample image, marked as y i . x1 i As the horizontal axis, y i As the vertical coordinates, these data points are placed in a two-dimensional coordinate system. These data points are approximately a straight line in the coordinate system. For the data point (x2 i ,y i ) also has a similar pattern. Let the number of people in the crowd be y, the area ratio of the foreground pixels in the crowd be x1, and the number of the edge pixels in the foreground of the crowd be x2, then we have formula (7):
[0061] Among them, a1, b1, a2, and b2 are line parameters.
[0062] The parameters of the two straight line equations can be calculated separately using the least squares fitting method. After performing corresponding image processing on the image to be tested to obtain the crowd target, the pixel area ratio and the number of edge pixels of the crowd foreground target are calculated, and then substituted into the curve equation to calculate the number of people in the image, thereby completing the estimation of crowd density based on the number of people.
[0063] Correspondingly, when the crowd density estimation is a crowd density estimation method based on texture features, the acquired image information is processed according to the determined crowd density estimation method, and the crowd density in the crowd density estimation area is estimated, including: S76 extracting texture features of the image information, the texture features include energy statistics, contrast statistics, entropy statistics, inverse moment and correlation; S77 inputting the extracted texture features into a pre-trained density classifier for classification, and completing the crowd density estimation in the crowd density estimation area; the density classifier is trained by the texture features and density classification levels of the training sample image.
[0064] Gray-level co-occurrence matrix can effectively describe the texture information of an image, and has the advantages of simple calculation, significant effect, and fast processing speed. The present invention uses the gray-level co-occurrence matrix to estimate the density of high-density crowds. The gray-level co-occurrence matrix is a matrix composed of the joint probability density P(i, j, d, θ) of pixel pairs, and its mathematical expression is defined as [P(i, j, d, θ)] N×N, where N represents the grayscale level, i.e., the grayscale value range of pixel pairs in the image. It can generally be 8, 16, 32, etc. The optimal grayscale level needs to be determined through experiments. d represents the distance, i.e., the number of pixels between pixel pairs. θ represents the direction, i.e., the relative angle between pixel pairs. Generally, only four angles are used: 0°, 45°, 90°, and 135°. i and j represent the grayscale values of the pixel pair, i, j = 0, 1, 2, ... N. P(i, j, d, θ) represents the value at position (i, j) in the gray-level co-occurrence matrix.
[0065] (1) Crowd density feature extraction
[0066] The present invention selects three characteristic quantities, namely energy statistics, contrast statistics and entropy statistics, as crowd density features. These three characteristic quantities are expressed as follows:
[0067] 1. Energy statistics En(d,θ), as shown in formula (8):
[0068] The energy statistic is the sum of the squares of all values in the calculation matrix, which mainly describes the uniformity of the grayscale distribution law of the image and the coarseness of the texture pattern.
[0069] 2. Contrast statistic Con(d,θ), as shown in formula (9):
[0070] Contrast statistics are used to measure the clarity of the image and the depth of the texture grooves.
[0071] 3. Entropy statistic Ent(d,θ), as shown in formula (10):
[0072] The entropy statistic is mainly used to describe the amount of information in an image. If the entropy value is close to zero, it means that there is almost no texture in the image; if the entropy value is large, it means that the image contains a lot of fine texture.
[0073] (2) Selection of gray-level co-occurrence matrix parameters
[0074] Since the gray-level co-occurrence matrix calculated using different parameters for the same crowd image will produce different results, the extracted crowd density features will also be different, resulting in different crowd density information. Therefore, it is necessary to select appropriate construction parameters to calculate the gray-level co-occurrence matrix so that the matrix can accurately and effectively reflect the texture information of the image.
[0075] 1. Selection of pixel grayscale parameters
[0076] Without reducing the accuracy of texture feature extraction, the image's grayscale levels are usually compressed to a certain extent to reduce the image's grayscale level, and then the grayscale co-occurrence matrix is calculated. This matrix has higher computational performance. In this embodiment, starting from a grayscale level of 16, the overall changes in the three statistical quantities of the image are roughly the same.
[0077] 2. Selection of pixel distance parameters
[0078] If the pixel distance is too large, the calculated gray-level co-occurrence matrix cannot accurately describe the texture information in the image; if the value is too small, the texture information described by the gray-level co-occurrence matrix is relatively complete, but the computational complexity is also greatly increased, and the differences between the texture features of different images are too small to perform effective classification. Therefore, in order to extract more accurate texture features, an appropriate pixel spacing must be selected to calculate the gray-level co-occurrence matrix. According to experiments, when the pixel distance is less than 5, the texture feature values of different types of samples are poorly differentiated, and the image cannot be effectively density-classified. As the pixel distance increases, the discrimination of the feature values also increases. However, when the distance exceeds a certain range, the discrimination of the feature values gradually decreases, and a large amount of effective texture information is still lost. In order to maximize the discrimination between texture feature values while retaining as much effective texture information as possible, this embodiment sets the pixel distance to 10.
[0079] 3. Selection of pixel direction parameters
[0080] Based on a set of 16 grayscale levels, the present invention calculates the curves of the change of texture features in the four directions of 0°, 45°, 90°, and 135°. According to experiments, the curves of the five texture features in the three directions of 0°, 45°, and 135° have basically the same change trends, while the curves of the 90° direction are more different from those in the other directions. Therefore, the present invention only selects the 0° and 90° directions to calculate the gray-level co-occurrence matrix.
[0081] (3) Support Vector Machine Classifier
[0082] Sample classification training is to select appropriate sample images from the collected video image sequence to obtain a sample set, then establish a gray-level co-occurrence matrix for the samples in the sample set and extract the texture features therein, and then send the extracted texture features into the classifier for classification training to obtain a crowd density classifier. The specific process is as follows:
[0083] 1. Select training sample images
[0084] To improve training accuracy and classification performance, the sample images selected must meet the conditions of comprehensiveness, sufficiency, and randomness. In accordance with these requirements, the present invention extracted 400 sample images of four density levels from experimental videos of the same monitoring environment at different time periods, with 100 sample images of each density level.
[0085] 2. Extract image texture features
[0086] First, each image in the selected sample set is compressed to 16 grayscale levels, and then the grayscale co-occurrence matrix in the two directions of 0° and 90° with a pixel distance of 10 is calculated respectively. Finally, according to the calculation formula of the feature quantity, the five texture feature quantities of the sample image, including energy, correlation, entropy, contrast and inverse moment, are calculated respectively. Among them, the grayscale co-occurrence matrix in the directions of 0° and 90° has a feature data respectively, and these 10 data of each sample image are combined into a 10-dimensional feature vector.
[0087] 3. Build a classifier
[0088] The extracted texture features are used to train the classifier. This embodiment uses a support vector machine (SVM) machine learning method, employing the RBF kernel function and a one-to-one multi-class classification method to perform classification training on the feature vectors of the dimension extracted from the sample image. Since there are five density levels in the sample image (as shown in Table 1), with a total of five classification categories, a total of seven density classifiers are trained.
[0089] The present embodiment is obtained by improving the above embodiment. In the present embodiment, the acquired image information is processed according to a determined hospital crowd density estimation method. After estimating the crowd density in the crowd density estimation area, a step of performing an early warning operation based on the assessed crowd density is also included.
[0090] In this embodiment, an early warning operation is performed based on the estimated results of the crowd density. According to the density level classification standard in Table 1, a multi-level early warning can be configured in actual applications. For example, when the density level is very high, an early warning prompt is issued by monitoring alarms, sending text messages, etc., and the hospital security personnel activate the early warning mode and initiate corresponding measures according to the early warning plan. When the density level is high, a reminder text message is sent to the hospital security personnel to pay attention to controlling the crowd density. When the crowd density is medium or below, it can be ignored. It should be clear that the above is only an example of an early warning method. In the application, the corresponding early warning operation can be configured according to the actual situation. No specific limitation is made here.
[0091] As shown in Figure 2, the present invention also provides a hospital crowd density estimation system, including: a camera device 10, which is configured in a crowd density estimation area and obtains image information of the crowd density estimation area; a hospital information system 20; a crowd density assessment terminal 30, which is connected to the camera device and the hospital information system respectively, and is used to determine the crowd density estimation area; obtain the number of people and the area of the crowd density estimation area based on the hospital information system; estimate the crowd density in the crowd density estimation area based on the number of people and the area; determine its density level based on the crowd density estimation in the crowd density estimation area; determine a crowd density estimation method based on the number of people and the density level in the crowd density estimation area, the crowd density estimation is a crowd density estimation method based on pixel features, or a crowd density estimation method based on texture features; and process the acquired image information according to the determined crowd density estimation method to estimate the crowd density in the crowd density estimation area.
[0092] With the rapid development of hospital information technology, hospital appointment registration systems, outpatient check-in systems, examination appointment systems, imaging or ultrasound triage systems, and ward bed management systems have become widespread. Hospitals have also generally established integrated service platforms, enabling data interconnection across these systems. These hospital information systems often contain real-time data from areas prone to congestion, such as outpatient waiting areas, examination waiting areas, and payment waiting areas. Based on this, this embodiment establishes a correspondence between surveillance images of the crowd density estimation area and real-world data from the hospital information system. This allows for a preliminary determination of the density level of the crowd density estimation area, and then selects a more efficient crowd density estimation method based on the determined density level, thereby improving the accuracy of crowd density estimation.
[0093] The crowd density estimation area can be a specific area within the hospital, such as the outpatient waiting area, the examination waiting area, the payment waiting area, and other areas prone to crowd congestion, or it can be an outpatient building and other areas. In the application, it can be determined based on the actual situation of the hospital.
[0094] The number of people in the crowd density estimation area can be the number of people in the crowd density estimation area obtained in real time, or it can be the average number of people in the crowd density estimation area at multiple moments within a pre-set period of time (such as 1 minute, 5 minutes, etc.). In the application, the method that best reflects the actual situation can be selected to determine the number of people in the crowd density estimation area based on the actual situation.
[0095] The area of the region is the area of the crowd density estimation area. This area can be determined based on the hospital's architectural conditions or pre-delineated and stored based on actual conditions. For example, a certain area at the payment window can be designated as the payment waiting area. Before estimating the crowd density, each crowd density estimation area and its corresponding area are determined and stored. This allows the corresponding area to be searched and read after a crowd density estimation area is selected during the crowd density estimation process.
[0096] In estimating the crowd density R, two methods such as formula (1) and (2) can be selected according to the situation. Regarding the selection of these two methods, if the number of people in the corresponding crowd density estimation area can be directly determined in the hospital information system, such as the outpatient waiting area, the examination waiting area, etc., the number of people in the corresponding area can be predicted based on the number of people signed in the system, and then the crowd density R can be directly calculated using formula (1). If the number of people in the corresponding crowd density estimation area cannot be directly determined based on the hospital information system, or the determination process is relatively complicated, such as the outpatient building, etc., the number of people inside can be directly determined based on the number of people entering and leaving the building (including patients and accompanying personnel, etc.), that is, the crowd density R can be determined using the method of formula (2). In order to obtain a more accurate number of people entering and leaving, the number of people can be counted in conjunction with the access control system.
[0097] Due to factors such as headcount and area, for example, some areas of a building, such as a waiting area, may be inaccessible to the public. Therefore, the two aforementioned methods also introduce a preset coefficient λ to regulate the calculation results. This preset coefficient λ needs to be determined based on actual conditions and is usually set between 0 and 1. In practical applications, since the area of the crowd density estimation area changes in real time, when estimating crowd density, it is often necessary to estimate the crowd density for a certain period of time, such as 1 minute or 5 minutes, to obtain the crowd density range for that period of time, which provides a more accurate basis for subsequent analysis.
[0098] In determining the density level based on the crowd density estimation in the crowd density estimation area, when the estimated crowd density is 0.8 people / m 2 If the number of people per 13 square meters is less than 10, the crowd density in the estimated area is low; when the estimated crowd density is higher than 0.8 people / m 2 Or if the number of people per 13 square meters is greater than 10, the crowd density in the estimated area is judged to be too high, as shown in Table 1.
[0099] Based on the above density level classification standards, when the crowd density in the crowd density estimation area is judged to be low, the crowd density estimation method based on pixel features is selected for estimation; when the crowd density in the crowd density estimation area is judged to be high, the crowd density estimation method based on texture features is selected for estimation.
[0100] In another embodiment, the hospital crowd density estimation system further includes an early warning device connected to the crowd density assessment terminal, for performing an early warning operation according to the assessed crowd density.
[0101] In this embodiment, an early warning operation is performed based on the estimated results of the crowd density. According to the density level classification standard in Table 1, a multi-level early warning can be configured in actual applications. For example, when the density level is very high, an early warning prompt is issued by monitoring alarms, sending text messages, etc., and the hospital security personnel activate the early warning mode and initiate corresponding measures according to the early warning plan. When the density level is high, a reminder text message is sent to the hospital security personnel to pay attention to controlling the crowd density. When the crowd density is medium or below, it can be ignored. It should be clear that the above is only an example of an early warning method. In the application, the corresponding early warning operation can be configured according to the actual situation. No specific limitation is made here.
[0102] It should be noted that the above embodiments can be freely combined as needed. The above are only preferred embodiments of the present invention. It should be pointed out that ordinary relevant personnel in this technical field can make several improvements and modifications without departing from the principles of the present invention. Such improvements and modifications should also be considered as the scope of protection of the present invention.
Claims
1. A method for estimating hospital population density, characterized in that: include: Determine the area of estimated population density; Obtaining the number of people and the area of the crowd density estimation area; estimating the crowd density in the crowd density estimation area based on the number of people and the area; Determining a density level of the crowd density estimation area based on the crowd density estimation; Determining a crowd density estimation method according to the number of people and the density level in the crowd density estimation area, wherein the crowd density estimation is a crowd density estimation method based on pixel features or a crowd density estimation method based on texture features; Acquiring image information of the crowd density estimation area; The acquired image information is processed according to a determined crowd density estimation method, and the crowd density in the crowd density estimation area is estimated.
2. The hospital population density estimation method according to claim 1, characterized in that: In estimating the crowd density in the crowd density estimation area based on the number of people and the area, the estimated crowd density R is: Among them, People is the number of people in the crowd density estimation area, m is the area of the area, and λ is the preset coefficient; Or, in estimating the crowd density in the crowd density estimation area based on the number of people and the area, the estimated crowd density R is: Among them, People1 is the number of people entering the crowd density estimation area, and People2 is the number of people leaving the crowd density estimation area.
3. The hospital population density estimation method according to claim 1 or 2, characterized in that: Determining the density level of the crowd density estimation area based on the crowd density estimation includes: When the estimated crowd density is 0.8 people / m 2 If the number of people per 13 square meters is less than 10, the crowd density in the crowd density estimation area is judged to be low; When the estimated crowd density is higher than 0.8 people / m 2 Or if the number of people per 13 square meters is greater than 10, it is determined that the crowd density in the crowd density estimation area is too high.
4. The hospital population density estimation method according to claim 1, characterized in that: When the crowd density estimation is a crowd density estimation method based on pixel features, the acquired image information is processed according to the determined crowd density estimation method, and the crowd density estimation in the crowd density estimation area includes: extracting a foreground image of the image information; Calculating a foreground pixel area of the foreground image; Performing edge detection on the foreground image and counting the number of edge pixels; Based on the least square method, linear models between the number of people in the crowd density estimation area and the foreground pixel area and the number of edge pixels are established respectively; The crowd density in the crowd density estimation area is estimated based on the linear module; Or, when the crowd density estimation is a crowd density estimation method based on texture features, the acquired image information is processed according to the determined crowd density estimation method, and the crowd density estimation in the crowd density estimation area includes: Extracting texture features of the image information, wherein the texture features include energy statistics, contrast statistics, entropy statistics, inverse moment and correlation; The extracted texture features are input into a pre-trained density classifier for classification, so as to complete the crowd density estimation in the crowd density estimation area; the density classifier is trained by the texture features and density classification levels of the training sample images.
5. The hospital population density estimation method according to claim 1, 2 or 4, characterized in that: The acquired image information is processed according to a determined crowd density estimation method, and after estimating the crowd density in the crowd density estimation area, a step of performing an early warning operation according to the estimated crowd density is also included.
6. A hospital crowd density estimation system, characterized in that: include: A camera device is disposed in a crowd density estimation area to obtain image information of the crowd density estimation area; Hospital Information Systems; The crowd density assessment terminal is connected to the shooting device and the hospital information system respectively, and is used to estimate the area according to the determined crowd density; obtain the number of people and the area of the crowd density estimation area based on the hospital information system; estimate the crowd density of the crowd density estimation area based on the number of people and the area of the area; determine the density level based on the estimated crowd density of the crowd density estimation area; determine the crowd density estimation method according to the number of people and the density level of the crowd density estimation area, and the crowd density estimation A crowd density estimation method based on pixel features or a crowd density estimation method based on texture features; and processing the acquired image information according to the determined crowd density estimation method to estimate the crowd density in the crowd density estimation area.
7. The hospital crowd density estimation system according to claim 6, characterized in that:The crowd density assessment terminal estimates the crowd density of the crowd density estimation area based on the number of people and the area, and the estimated crowd density R is: Among them, People is the number of waiting people in the crowd density estimation area, m is the area of the area, and λ is the preset coefficient.
8. The hospital crowd density estimation system according to claim 6, characterized in that: The hospital crowd density estimation system also includes an access control system connected to the crowd density assessment terminal for counting the number of people entering and leaving the crowd density estimation area; The crowd density assessment terminal estimates the crowd density of the crowd density estimation area based on the number of people and the area, and the estimated crowd density R is: Among them, People1 is the number of people entering the crowd density estimation area, and People2 is the number of people leaving the crowd density estimation area.
9. The hospital crowd density estimation system according to claim 6, 7 or 8, characterized in that: The crowd density assessment terminal determines the density level of the crowd density estimation area based on the crowd density estimation, including: When the estimated crowd density is within 0.8 or the number of people is less than 10, it is determined that the crowd density in the crowd density estimation area is low; When the estimated crowd density is higher than 0.8 or the number of people is greater than 10, it is determined that the crowd density in the crowd density estimation area is too high.
10. The hospital crowd density estimation system according to claim 6, 7 or 8, characterized in that: The hospital crowd density estimation system also includes an early warning device connected to the crowd density assessment terminal, which is used to perform early warning operations according to the assessed crowd density.
Citation Information
Patent Citations
Method and device for estimating crowd density in video image
CN101727573A
Crowd density estimation method, device and monitoring system
CN101739569A
Self-intended crowd density estimation method for camera capable of straddling
CN102982341A
Adaptive abnormal crowd behavior analysis method
CN103745230A
Surveillance video transmission method based on video content analysis
CN105791774A
