JP2012160165A - Baseline band video monitoring system and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To promptly identify a moving locus of a vehicle with higher accuracy and to calculate traffic information in a road section to be monitored in real time.SOLUTION: A baseline band video monitoring system includes baseline band preprocessing means configured to perform preprocessing on an image of a view field to be monitored captured by an imaging apparatus, thereby acquiring one or more baseline bands, feature point extraction means configured to perform feature point extraction on a moving object within the baseline band, thereby acquiring a binarized contour of the moving object present in any one of the one or more baseline bands and a feature pixel point, and moving object identification means configured to collate the binarized contour of the moving object in the baseline band and the feature pixel point between a plurality of images, thereby identifying and detecting the moving object.

Description

  The present invention relates generally to the field of computer software information processing, and more particularly to the recognition of video image patterns used in intelligent communications.

  In recent years, with the rapid increase in the number of moving vehicles accompanying economic development, road traffic conditions in most cities have been deteriorating, so it is an urgent issue to improve the control ability for road traffic in an intelligent way. It has become. In the control of road traffic, the acquisition of traffic conditions is one of the most important factors. In general, traffic state information is acquired by conventional methods such as electromagnetic waves and infrared rays. However, conventional devices are fragile and not easy to repair. In order to cope with this problem, a method called video surveillance is applied, and this method is becoming an intelligent and effective communication control mode. In general, the video image monitoring method is performed by a procedure in which a video data stream is captured by a camera installed above a road, vehicle movement information is extracted from the video data stream, and real-time traffic information is calculated. Is called. In this method, it is important to monitor and analyze road traffic and communication flow quickly and accurately.

  FIG. 1 is a processing flow 100 for processing real-time traffic information in an existing video surveillance system. In step S012, when the vehicle enters the field of view field monitored by the camera, the camera first identifies the contour image information of the vehicle (vehicle body), and uses the identified contour image information to processing means (not shown). Send in the background. As this processing means, a general-purpose or dedicated processing device such as a CPU or ASIC can be used. In step S104, the processing means processes the contour image information of the vehicle. Specifically, the processing means identifies the position (generally, the front part and / or the rear part of the vehicle) where the license plate is attached from the contour image, and extracts the license plate contour area information. The technique for identifying the license plate mounting position and extracting the license plate outline area information is well known in the art and will not be described in detail here. Next, in step S106, the processing means processes the license plate area information to further extract character information (vehicle number information) in the license plate area. In this case, there is generally a difference in angle between the camera installation position and the moving direction of the vehicle, and the extracted license plate outline is often not a front image of the license plate (typically It is difficult to extract the character information of the license plate. Therefore, the processing means needs to correct the rotation angle of the acquired license plate so that the image of the license plate becomes a front image. As an example of the method, there is a method of determining whether the image obtained after the rotation is a front image after rotating the contour region information of the license plate over a preset angle. If the result is “yes”, the process ends. If the result is not “yes”, the processing means adaptively adjusts the rotation angle (for example, by adding or subtracting a constant increment value), rotates the image, determines again, and the front of the license plate This is repeated until an image is obtained. Techniques for determining by rotating image information are also well known in the art and will not be described in detail here. Furthermore, a technique for extracting character information from image information in the front direction is also well known.

  Through the processing in step S106, the processing means acquires the character information (vehicle number) of the license plate that can uniquely identify the vehicle as the vehicle monitoring flag. This monitoring flag makes it possible to easily identify the position when the vehicle to be monitored further moves.

  Note that the processing of steps S102 to S106 is performed on one image, but in actual use, the video stream is continuously input from the camera to the processing means. Therefore, in step S108, the processing unit extracts image information of the entire visual field from one image group included in the continuous video stream, and executes the processes of steps S102 to S106 on the image group. Thus, the vehicle to be monitored is detected and the position is specified, and the movement trajectory of the vehicle in the field of view of the camera is identified based on the detected position.

  Next, in step S110, the processing means calculates the moving distance of the vehicle based on the movement trajectory, and then calculates the moving speed of the vehicle based on the time required for the moving distance in step S112. Get traffic information for lanes.

  With the above processing, it is possible to obtain the moving speed of the vehicle in the visual field as a monitoring target, and it is also possible to derive complex traffic information from the speed information of a plurality of vehicles in the visual field.

  However, this processing method has the following drawbacks. First, it is known in the technical field that calculation for extracting character information from an image is time consuming and difficult. That is, the process of extracting the license plate outline from the vehicle outline information and further extracting the license plate character information is complicated, and takes a very long time especially when there are many vehicles in the field of view. . Furthermore, in order to detect the vehicle to be monitored and specify the position, it is necessary to process each vehicle in the field of view and obtain the character information of each license plate, thus guaranteeing real-time execution of the processing. It is almost impossible to do. In addition, when the angle difference between the camera installation position and the moving direction of the vehicle is large, it is difficult to obtain a front image of the license plate at that time by correction. In such a case, since there is a possibility that extraction of the character information of the license plate may fail, the monitoring accuracy is lowered.

  The second disadvantage is that when the monitoring range of the visual field is large, the size of the acquired image is large, so that the vehicle contour information included in the image is very large. In this case, the calculation for extracting character information from the image for tracking increases, and the amount of calculation becomes particularly large in the case of a series of continuous images. Therefore, real-time execution of processing becomes extremely difficult.

  Therefore, with existing road traffic monitoring methods, it is difficult to guarantee the execution and accuracy of processing in real time, especially when the field of view field to be monitored is wide and there are many vehicles in the field of view. Is almost impossible. Therefore, it is an urgent issue to provide a rapid and highly accurate road traffic monitoring method that can quickly and more accurately identify the movement trajectory of a vehicle and that can calculate traffic information of a monitored road section in real time. .

  According to one aspect of the present invention, baseline band preprocessing configured to preprocess an image of a monitored field of view captured by an imaging device, thereby obtaining one or more baseline bands. And obtaining a binarized contour and feature pixel points of the moving object existing in any one or more of the baseline bands by performing feature point extraction on the moving object in the baseline band The feature point extracting unit configured to identify and detect the moving object by performing the matching of the binary contour and the feature pixel point of the moving object in the baseline band between the plurality of images. A baseline band video surveillance system comprising a moving object identification means is provided.

  According to another aspect of the invention, pre-processing an image of a monitored field of view captured by an imaging device, thereby obtaining one or more baseline bands, and any of the baseline bands , Obtaining a binarized contour and feature pixel points of the moving object by performing feature point extraction on the moving object in the baseline band, and binary of the moving object in the baseline band A baseline band video surveillance method is provided comprising: identifying and detecting moving objects by performing a match between a plurality of images of normalized contours and feature pixel points.

  In the present invention, at least the following points are improved.

  1. Since the vehicle can only move on the road, the road or lane indicated in the field of view to be monitored is pre-labeled in a band as a baseline band used as a monitoring range. As a result, it is not necessary to process a large image including the entire field of view, so that the monitoring range can be narrowed and the amount of calculation of the system can be reduced.

  2. At the time of identifying the vehicle contour, the color of the vehicle in the baseline band is greatly different from the background color of the road, so the binarization process is executed on the original image to form a black and white image. As a result, the amount of calculation is greatly reduced compared to the original color image, so that the contour information of the vehicle can be identified more quickly.

  3. The characteristic pixel points of the vehicle are extracted from the contour region of the vehicle within the baseline band, but at this time, there is a high possibility that there is license plate information that can more accurately identify the vehicle at the front and rear of the vehicle. The front and rear sampling needs to be performed at a higher density. A vehicle is uniquely identified by contour information and feature pixel points. Thereby, the complicated calculation for recognizing the character of a vehicle number can be avoided, and calculation efficiency can be improved. Furthermore, since the failure of character recognition due to the camera installation angle or the like can be avoided, the identification accuracy is improved.

  Therefore, in the present invention, at least as compared with the related art, 1) the monitoring range is divided into a plurality of baseline bands based on lanes or roads, and the monitoring range of the system is narrowed. Since the contour can be quickly identified, the processing performance is improved. 3) Since the vehicle identification information is acquired by extracting the characteristic pixel points, the processing speed and accuracy in a specific situation are greatly improved. Benefits can be gained.

A processing flow 100 for processing real-time traffic information in an existing video surveillance system. 2 is a functional block diagram of a multiple baseline band video surveillance system 200 according to the present invention. FIG. 3 is an operational flow 300 of a multiple baseline band video surveillance system according to the present invention. A plurality of baseline bands obtained by the processing of the plurality of baseline band preprocessing means 201. It is a detailed block diagram of the feature point extraction means 202 and the vehicle identification means 203 of this invention. It is the result of the binarization process by the binarization means 504. It is a figure which shows the sampling operation | movement with respect to a vehicle based on a sampling rule. It is a detailed operation | movement flow of step S306-S308 of FIG. It is an example of the characteristic information collation process in step S802 of FIG. It is a figure which shows the determination of monitoring effectiveness. It is a figure which shows the adaptive adjustment of the range of a baseline band.

  Next, preferred embodiments of the present invention will be described with reference to the drawings.

  FIG. 2 is a functional block diagram of a multiple baseline band video surveillance system according to one embodiment of the present invention. The multiple baseline bandwidth video monitoring system 200 includes multiple baseline bandwidth preprocessing means 201, feature point extraction means 202, and vehicle identification means 203. Further, the video surveillance system 200 may further include a speed calculation unit 204 and an analysis unit 205 as necessary. In the following, as an example of a multiple baseline band video surveillance system 200, a system configured to analyze lane road traffic conditions to obtain road traffic information will be described. However, it should be noted that the monitoring system 200 is not limited to a system configured to analyze lanes, and not all means need be included in the system 200. In use environments other than roads, it is possible to perform moving object monitoring using all or a part of the system. For example, the monitoring system 200 can be used as a human traffic monitoring system for analyzing human traffic information in a subway station. In the following, for convenience of explanation, a vehicle on a lane is used as an example of a moving object.

  The plurality of baseline band pre-processing means 201 has one or more existing in the field of view field to be monitored in order to acquire one or more baseline bands each having a monitoring start line and a monitoring end line as a monitoring unit. Are pre-processed and one or more baseline bands (hereinafter referred to as “multiple baseline bands” in the description below) are included in the feature point extraction means 202, including the case where there is only one baseline band. Output). The monitoring start line and the monitoring end line of one baseline band may be fixed during monitoring or may be dynamically adjusted by the system 200.

  The feature point extraction unit 202 is configured to perform feature point extraction on a plurality of images of the vehicle in the baseline band, and to acquire a binarized contour of the vehicle and one or more feature pixel points. . The vehicle identification unit 203 is configured to execute the binarized contour of the vehicle and the feature information collation of the feature pixel points between a plurality of images to acquire the trajectory of the vehicle within the baseline band. The acquired trajectory is used by the speed calculation means 204 to calculate the vehicle speed. The analysis means 205 is configured to calculate the road traffic state based on the vehicle speed and the traffic state output. Hereinafter, specific configurations of the feature point extraction unit 202, the vehicle identification unit 203, and the speed calculation unit 204 will be described in detail with reference to FIG.

  In addition, with reference to FIG. 3, the operation of each unit of the multiple baseline band video monitoring system 200 will be described in detail. FIG. 3 shows an operational flow 300 of the multiple baseline band video surveillance system 200 of the present invention.

  In step S302 of FIG. 3, the multiple baseline band preprocessing unit 201 first preprocesses the field-of-view field to be monitored to obtain a multiple baseline band. Specifically, since the vehicle can move only on the lane, the multiple baseline band pre-processing means 201 divides and marks the range of the visual field to be monitored in a band shape, and based on the movement direction rule of the entire lane, A specific line corresponding to a position where the vehicle enters the range of the baseline band is set as a monitoring start line of the baseline band, and another line corresponding to a position where the vehicle leaves the range of the baseline band is set as the monitoring end line of the baseline band. Set as. FIG. 4 shows a plurality of baseline bands obtained by the processing of the plurality of baseline band preprocessing means 201. It is noted that the method of dividing the lane in the field of view to be monitored and the method of setting the lane monitoring start line and the monitoring end line are arbitrary, and may be either artificial recognition or automatic recognition by a machine. I want. The multiple baseline band preprocessing means 201 extracts feature points of information on multiple baseline bands (including information on start lines and end lines of individual baseline bands) obtained by division based on rules determined in advance. It outputs to the means 202.

  In step S304, the feature point extraction unit 202 is input based on the information on the multiple baseline bands transmitted from the multiple baseline band pre-processing unit 201 (including information on the monitoring start line and the monitoring end line). Feature point extraction is performed on the continuous traffic video stream, thereby obtaining a binarized contour of the vehicle and sampling to derive a feature pixel point of the vehicle.

  In step S306, the vehicle identification unit 203 identifies and detects the vehicle as a monitoring target by executing vehicle collation between a plurality of images based on the provided feature information.

  In step S308, the speed calculation unit 204 calculates a movement trajectory and a movement time based on the result of vehicle collation between a plurality of images, and calculates a movement speed of the vehicle based on the calculated movement trajectory and the movement time. To do. Next, a detailed configuration and function of the speed calculation unit 204 will be described with reference to FIG.

  In step S310, the analysis unit 205 calculates an average speed of a plurality of vehicles moving on a specific lane in a recent period (for example, 3 minutes), and outputs the result as a result of road traffic information for the period. It should be noted that the setting of this period (eg, 3 minutes) is merely an example, and any appropriate time required as a period for evaluating road traffic information can be selected. For example, when it is necessary to monitor a lane to be monitored with higher accuracy, an average vehicle speed in a shorter period (eg, 30 seconds) may be selected as road traffic information of the lane. In another example, in order to obtain road traffic information, various vehicle speeds can be weighted and averaged based on different weighting factors.

  It should be noted that the functional structure of the multiple baseline band video surveillance system 200 and the operational flow 300 of the surveillance system 200 described above are only examples. Therefore, a person skilled in the art can rearrange the functions of the respective means as necessary to combine the functions of several means or further divide the functions of several means. For example, the functions of the vehicle identification unit 203 and the speed calculation unit 204 can be combined into one unit, or the functions of the speed calculation unit 204 and the analysis unit 205 can be combined into one unit. It is also possible to combine the functions of the three means into one means. The means may be configured to perform vehicle verification between a plurality of images, calculate a movement trajectory and a vehicle speed based on the verification result, and obtain road traffic information based on the vehicle movement speed. it can.

  In the functional block diagram of FIG. 2, some well-known components are not shown. For example, the video surveillance system 200 includes, of course, a video camera for capturing images and a buffer memory for storing traffic video streams received for subsequent feature point extraction and matching operations. However, since the video camera and the buffer memory are not directly related to the present invention, these components are not shown and are omitted from the drawing.

  FIG. 5 is a detailed configuration diagram of the feature point extraction unit 202, the vehicle identification unit 203, and the speed calculation unit 204 according to an embodiment of the present invention. This figure is only an example, and as described above, those skilled in the art can combine or further divide the functions of the respective means as necessary.

  The feature point extraction unit 202 shown in FIG. 5 includes an image change portion identification unit 502, a binarization unit 504, and a feature pixel point sampling unit 506. The speed calculation unit 204 includes a trajectory drawing unit 510 and a vehicle speed calculation unit 512. Below, the function of each means will be described in detail.

  The video stream received by the feature point extraction unit 202 is a continuous video stream captured at a constant frame rate (usually 24 frames / second, but may be other frame rates if necessary). . Therefore, after the video stream is received, the image change portion identifying unit 502 first acquires static image information at the current time point, and based on the static image information, pixel information in a specific range near the start line of the baseline band Is compared with the pixel information at the previous time point (here, the image of the current frame is compared with the image of the previous frame. The comparison object is pixel information in a specific range near the start line of the baseline band) Then, the change part is identified based on the comparison result. This specific range (specific range) needs to be set longer than the vehicle length so that the entire vehicle completely falls within this range (region). Hereinafter, this range is referred to as a “start line monitoring range”. In other words, the presence of the change portion indicates that the vehicle is in the process of entering the baseline band. Specifically, when the vehicle is traveling on a road, there is some distance between the two vehicles, so pixel information of the start line of the baseline band (that is, the start line of the monitoring range) If the pixel information in a nearby small area) does not change and the pixel information in the area other than this area in the monitored area changes, the vehicle is in the process of entering the baseline band and the vehicle body is Since it indicates that the area has been completely entered, the monitoring process can be started at this point. On the other hand, when the vehicle is in the process of entering the monitoring start line of the baseline band (for example, when the image change part identifying unit 502 identifies the change part in the image, 30% of the vehicle enters the baseline band. Note that if this vehicle is covering the baseline start of monitoring in the baseline band, the pixel information near the start line will also change. In this case, since the vehicle is not completely within the baseline band, the trigger condition for starting monitoring is not satisfied. Thus, the monitoring process is initiated only when the vehicle is fully within the baseline band.

  Next, the image change portion identification unit 502 notifies the binarization unit 504 and the characteristic pixel point sampling unit 506 about the change portion for subsequent processing. Here, it should be noted that the “previous frame image” does not necessarily have to be a frame image immediately before the current image, and may be an image two or four before the current image. How much the previous image is acceptable depends on the vehicle speed, the road traffic state in the previous processing period, the expected accuracy, and the like, as will be described later.

  In addition, the above “specific range (region) near the start line” is a set range that ensures that a vehicle that falls within the monitored baseline band can be identified (that is, not ignored by the monitoring system). As long as it is (area), the system administrator may set in advance based on, for example, the average speed (or speed limit, speed at the previous time, etc.) of the monitored lane (baseline band). Further, the setting range (area) can be adaptively changed during the monitoring procedure. Furthermore, the method by which the image change portion identification unit 502 identifies the change portion is not limited to the comparison of pixel information in a range (region) near the start line of the baseline band, and other methods are also possible. For example, it is possible to set several points near the start line in advance and monitor changes only at these points, which greatly reduces the amount of comparison calculation and improves real-time characteristics. Can be made. Further, it may not be necessary to provide the image change portion identifying means 502. For example, it is possible to cause the system to perform binarization processing and feature point sampling operation on an image captured at a certain interval without determining when the vehicle crosses the monitoring start line. For example, when the frame rate is 24 frames per second as described above, it can be set every other frame or every other frame). However, if the image change portion identifying means 502 is omitted, trouble may occur. For example, if the binarization and sampling operation (and subsequent operations) are executed during the period when the vehicle does not enter the lane for a long period of time, the power consumption increases unnecessarily. In such a case, the power consumption can be reduced by providing the image change portion identifying means 502.

  The image change part identification unit 502 notifies information related to the change of the image (that is, information that the current image in the specific range (region) near the start line is changed compared to the immediately preceding frame) Receiving the notification, the binarization means 504 extracts vehicle contour information in the baseline band by binarization processing. Specifically, since the color of the vehicle body and the color of the road background are generally different, the image of a specific range (area) in the vicinity of the start line of the baseline band is converted into a black and white image, and then the vehicle body pixels that differ greatly from the background It is possible to execute a binarization process on the features. FIG. 6 shows the binarization result. Although FIG. 6 shows a case where a vehicle whose body is completely within the baseline band is binarized, even if a plurality of vehicles are simultaneously present in the baseline band, the same principle is applied to perform binarization. Can be performed. In addition, it is possible to monitor only one vehicle within the same baseline band. That is, when one vehicle enters a specific range within the monitoring start line of the baseline band, the monitoring process is started, and until the vehicle moves out of the monitoring end line or until the vehicle moves from the monitoring end line. Other vehicles that have entered the baseline band are not monitored until they are somewhere without going outside (for example, if the vehicle changes lanes or leaves this baseline band). In this case, since the maximum number of vehicles monitored within one unit time is one, the system load is reduced. However, since the number of vehicles monitored within one unit time also decreases at the same time, the monitoring accuracy decreases.

  Further, in the drawing of FIG. 6, the pixel of the background portion of the lane is “1” and the pixel of the contour of the vehicle body is “0”, but this is only an example, and the pixel of the background portion of the lane is “0”. It is also possible to reverse the body contour pixels, such as “1”. Also, any other notation can be used as long as the vehicle body contour and the lane background can be distinguished.

  By the processing of the binarizing means 504, vehicle contour information (that is, contour information of one or more vehicles in the image) is acquired. In other words, the first color image captured by the video camera becomes a black and white image by the processing of the binarizing means 504, and binarized contour information can be acquired. By this processing, the amount of information to be processed is greatly reduced compared to the first color image (for example, an image of 256 grade).

  Next, the feature pixel point sampling means 506 performs a feature point sampling process on the vehicle pixel points included in the region identified by the vehicle contour in the first color image of the baseline band. In this feature point sampling process, it is possible to reduce the calculation amount by thinning out (decreasing) the pixel points of the vehicle. The principle of thinning operation is, for example, thinning out pixel points that are significantly different in color from surrounding pixel points in a specific range, or thinning out pixel points so that the color difference between multiple pixel points in a specific range is small There are multiple types, such as thinning out pixel points so that a representative one of the pixel points is selected. In any case, the decimation operation must be performed to ensure that the vehicle is uniquely identified. In addition, the principle of the thinning operation used must be the same in the same processing unit. This is because when comparing sampling points in two sampling point sets, each sampling point to be compared in the two sampling point sets must be at the same or adjacent location on the vehicle. In the actual sampling operation, since there is pixel point information of the license plate that can more uniquely identify the vehicle at the front and rear of the vehicle, it is necessary to sample the front and rear of the vehicle at a higher density. is there. In contrast, when sampling other parts of the vehicle, the sampling points may be relatively sparse. For example, when sampling the uppermost part of the vehicle body (it is not easy to distinguish the vehicle at the uppermost part of the vehicle body), the number of sampling points can be small or zero.

  However, the pixel point information of the vehicle license plate is only an example desirable for uniquely identifying the vehicle. In fact, the information for uniquely identifying the vehicle is not limited to the pixel point information of the vehicle license plate. For example, in some cases, a vehicle may be uniquely identified using a combination of a model (type), a color of the vehicle, an ID attached to the front window, and the like. In addition, the system administrator can set the sampling density of different parts based on specific conditions, or can be set automatically. FIG. 7 is a diagram illustrating a sampling operation for the vehicle based on the sampling rule.

  Feature pixel point information for uniquely identifying the vehicle is acquired by the processing of the feature pixel point sampling unit 506. Based on the binarization result of the binarization unit 504, the characteristic pixel point sampling unit 506 performs the characteristic pixel point sampling process on the pixel points in the area identified by the vehicle body contour, and uniquely identifies the vehicle. It is configured to acquire feature pixel point information that can be obtained. The amount of information of characteristic pixel points acquired by this method is greatly reduced compared to the first image of the entire color vehicle body.

  Through the above processing, the feature point extraction unit 202 acquires vehicle feature information, that is, binarized contours and feature pixel points, and supplies this feature information to the vehicle identification unit 203.

  The vehicle identification unit 203 performs a feature information matching process based on the supplied binarized contour and feature pixel points of a plurality of images of a continuous video stream, and is a monitoring target in the baseline band. Get the verification result of the vehicle. Thereafter, the vehicle identification unit 203 identifies (detects) a vehicle to be monitored in a plurality of images based on the result of vehicle comparison.

  Next, the trajectory drawing means 510 of the speed calculating means 204 calculates the movement trajectory of the vehicle based on the identification result by the vehicle identifying means 203. Thereafter, the vehicle speed calculation means 512 calculates the actual road travel distance based on the travel trajectory of the vehicle, and acquires the travel speed of the vehicle. The vehicle speed calculation means 512 outputs the acquired moving speed to the analysis means 205 in order to obtain the road traffic state. Next, operations of the vehicle identification unit 203 and the speed calculation unit 204 will be described with reference to FIG.

  FIG. 8 is a detailed operation flow of steps S306 to S308 (that is, vehicle identification and speed calculation).

  In step S802 in FIG. 8, first, the vehicle identification unit 203 performs feature information collation to identify (detect) the vehicle. Specifically, the video stream input from the camera is refreshed in the form of a static image (frame). Since the frequency of refreshing video frames per second is generally high (usually 24 frames / second), there is very little vehicle displacement in adjacent static images or static images that are several frames apart. For example, even if the moving speed of the vehicle is 72 km / hour (however, this is a speed that cannot be achieved on urban roads), the displacement per second is only 20 m, so the relative displacement of the same vehicle in each frame Is less than 1 m.

  Therefore, as described above, when detecting the range of the vehicle position, based on the baseline band position of the vehicle in the previous image, the change feature with respect to the vehicle image in a narrow range along the predetermined movement direction of the lane Extraction can be performed. Thereafter, feature information matching is performed on the feature information (contour information and feature sampling points) of the vehicle between the previous image and the current static image. This “narrow range” may be set by the system administrator based on actual conditions, or may be set automatically, or may be a fixed value, or may be included in road traffic information in the previous period. The value may be adjusted adaptively based on the value. Furthermore, when a part of the lane is hidden behind a building or tree, the value of the range for detecting the vehicle position can be increased according to the way of hiding. In the matching process, it is not necessary to perform comparison and matching for all frames, and the matching may be executed every two frames, every four frames, or every second. For example, when the vehicle is moving at a high speed, it is necessary to perform comparison and collation for every frame, but when there is a traffic jam (in this case, the vehicle speed is 30 km / hour or less. There is no need to do this. As described above, the interval of the matching process can be adaptively adjusted according to the traffic information acquired immediately before.

  FIG. 9 shows an example of the feature information matching process. Hereinafter, this process will be described. First, in step S902, vehicle contour information is collated. In the area defined by the contour information of the vehicle, when the outer shape of the vehicle changes, the current vehicle verification is stopped and the verification process is terminated. For example, in this collation, the passenger car and the microbus are easily distinguished and identified because they are clearly different in outer shape and size, and are determined as different vehicles. Furthermore, in the contour information matching process, the vehicle may change the moving direction (moving angle). In this case, fine adjustment and correction can be performed by rotating the vehicle contour. If the contour information matches, the color of the feature sampling point in the vehicle contour is collated in step S904, and if the color difference is large, collation of the current vehicle is stopped. By comparing in step S904, two passenger cars having similar external shapes and different colors can be distinguished and identified. If the colors also match, the process proceeds to step S906. In step S906, further verification is performed based on the detailed information of the sampling points. Examples of this detailed information include the details of the front part of the vehicle, the details of the rear part of the vehicle, the details of the front window and the rear window of the vehicle. In the preferred embodiment, the detailed verification comprises a vehicle license plate area verification step. In other words, if two passenger cars are traveling in the front and back of the lane, and the outline and color of the two passenger cars are the same (for example, both are black, the same brand, the same model passenger car), The passenger car details are further compared (often in the vehicle license plate area). Even if the outline, color and details of two different passenger cars are exactly the same (for example, the vehicle numbers of the two passenger cars are the same except for a small difference. If this is not possible, or if the two passenger cars have the same vehicle number, are close to each other, and are traveling on the same lane), this possibility is very low and can be ignored. In conclusion, if the three elements (model (contour), color, and detail) match in the above collation, in step S908, the comparison target vehicles in the two images are the same vehicle (that is, the collation results match). ) Can be determined. In addition, there is a possibility that the color of the two frames may be different depending on the light beam, weather, etc., so a certain amount of margin is required in the color matching process and the detailed matching process. That is, in the comparison of feature pixel points, a tolerance of a specific threshold range is assumed. (This threshold can be set in advance or it can be adjusted as needed). However, in general, there is almost no difference between the colors of two frames acquired at a slight time interval. In addition, in the above collation processing, collation is executed in the order of model, color, and detail, and if each step does not match, the next collation is stopped. This may greatly reduce the amount of calculation. It will be appreciated, however, that the collation need not necessarily be in this order, can be performed in a different or reverse order, and one or more elements of the collation can be added or removed. Therefore, the verification target may be any one or more of the vehicle model, color, and details, or a combination thereof. For example, because the amount of detailed matching is the largest, only model matching and color matching can be performed to improve real-time performance, but it goes without saying that this sacrifices system accuracy. Alternatively, if only the detailed matching is performed and the matching of the contour and the color is omitted, the operation procedure can be reduced, but the calculation amount is unnecessarily increased. This is because if two vehicles in the line band have clearly different colors, the subsequent verification is not necessary and the verification process can be terminated at this stage). Further, only the vehicle license plate region is verified by the detailed verification, and the front portion of the vehicle, the window portion, etc. can be omitted. A person skilled in the art can freely make this setting based on the accuracy actually required. Furthermore, in matching colors and details, one skilled in the art can define measurable principles of construction as needed. For example, when comparing the sampling points on the vehicle body or the detailed sampling points, if the percentage of the matching feature points in the entire feature points falls below the threshold value, the match is judged to be insufficient or failed. The subsequent operation may be omitted.

  Returning to FIG. 8, a series of vehicle position verification results in the baseline band are acquired by the process in step S <b> 802, and the monitoring target vehicle is identified. Next, in step S804, the trajectory drawing unit 510 executes a movement trajectory calculation process. Specifically, the movement trajectory of the vehicle can be obtained by connecting several positions of the same vehicle in the baseline band based on the acquired collation result. Then, based on the trajectory information, the pixel distance in the baseline band is further obtained by obtaining a start time when the vehicle enters the baseline band and an end time when the vehicle leaves the baseline band. Can be obtained.

  Next, in step S806, the vehicle speed calculation means 512 calculates the movement distance and the movement speed. As described above, the result obtained in step S804 is the pixel distance, but the pixel distance is not the same as the actual distance on the road. Therefore, in step S806, the vehicle speed calculation unit 512 converts the pixel distance into the actual travel distance on the road, and calculates the travel speed of the vehicle based on the travel distance and the travel time. Here, it is noted that the actual distance between two lines on the road is determined in advance when the multiple baseline band pre-processing means 201 has set the monitoring start line and the monitoring end line in advance. I want. Therefore, in this case, it is easy to convert the pixel distance into a moving distance on an actual road.

  Next, the above process will be further described by taking the following case as an example. Assume that three passenger cars A, B, and C are running vertically in the baseline band at the current time. The binarizing means 504 acquires the contour information of the three passenger cars, and the characteristic pixel point sampling means 506 executes characteristic pixel point sampling processing for these passenger cars. By this sampling, characteristic pixel point information of three passenger cars is obtained.

  Next, the vehicle identification unit 203 uses the binarized contour information and feature sampling point information of the current image as the previous image at the previous time point (for example, 1/12 seconds before or 1/6 second before the passenger car). Compared with the binarized contour information and feature sampling point information of the image including only A and passenger car B), passenger cars A and B exist in the previous image, and passenger car C newly enters the baseline band. Confirm that you have come. In this process, it is assumed that the passenger cars A and B belong to the same model (for example, Toyota Yaris (Vitz's European specification)) but have a different color, and the passenger car C has a car model that is clearly different from the passenger cars A and B. Accordingly, in step S902, C does not match any passenger car in the previous image, so it is determined that the car is a newly entered car. At the same time, the comparison between A and B proceeds to step S904 (color verification), where the passenger car A in the current image matches the passenger car (A) in the previous image, and the passenger car B is also the passenger car (B) in the previous image. Is determined to match. In this case, the movement locus and movement distance of the passenger car A are obtained, and the movement speed of the passenger car A from the previous time point to the current time point is obtained. Similarly, the moving speed of the passenger car B from the previous time point to the current time point is also obtained.

  As another example, suppose C belongs to the same car model as one of A and B, and the colors are different for both A and B. In this case, only by the process of step S904, it is determined that C does not match any passenger car in the previous image. As another example, suppose C belongs to the same car model as one of A and B, and the color is the same as one of A and B. In this case, in order to determine that C does not match any passenger car in the previous image, all three types of matching processes are required. As described above, when three matching processes (model, color, and detail) are executed, the possibility of erroneous determination is extremely low.

  By the above collation, the trajectories of the passenger cars A and B from the previous time point to the current time point are obtained. And time passes and it advances to the next time. As before, images at this time are acquired, and binarized contour information and feature pixel point information of three passenger cars are obtained. Thereafter, this information is compared with the binarized contour information and feature pixel point information of the image at a time point earlier than this time point (ie, the “current time point” described above), so that the passenger cars A and B at this time point are compared. , C are determined to coincide with the passenger cars A, B, C at the previous time points, respectively. In this case as well, the movement trajectory and movement distance of each of the passenger cars A, B, and C can be obtained, and the movement speed of each passenger car is calculated based on the movement locus and the movement distance. Next, the speed calculation means 204 synthesizes the average moving speeds of the respective passenger cars obtained during a certain period (for example, 3 minutes), thereby acquiring traffic information during that period.

  In the above example, video surveillance system 200 calculated vehicle speed each time a comparison was performed (ie, each time a trajectory was acquired), this is not necessary. If the speed is calculated after two or more trajectories are acquired, the amount of calculation can be reduced. For this purpose, the video surveillance system 200 may be set to calculate the vehicle speed after a series of vehicle matching in a plurality of images is completed. As described above, those skilled in the art can set this as necessary.

  In the above description, a case where a new vehicle has entered beyond the monitoring start line of the baseline band at a certain point is taken as an example, but the present invention is not limited to this. Actually, the vehicles in the baseline band are lined up continuously, and flow lines are drawn into the baseline band. Correspondingly, the video surveillance system 200 has the ability to continuously perform image capture, feature point extraction, speed calculation, and road traffic analysis operations. As described above, in order to reduce the amount of calculation and improve the real-time performance, the interval for performing feature point extraction and speed calculation for the input video stream is adaptively adjusted according to the actual road traffic condition. It is possible. For example, the time interval can be extended to 1/6 second or 1 second.

  In the case of the movement trajectory and time calculation described above, there is a problem that the verification in the verification process fails when a vehicle in the lane leaves the monitoring range of the baseline band (for example, due to a lane change). Please be careful. In such a case, it is necessary to determine the effectiveness of monitoring.

  A vehicle enters the baseline band at time T1, and at time T2, monitoring of the vehicle is not possible in the baseline band (because the lane has been changed to another baseline band). In order to cope with this problem, a variable ValidDis = S1 / S can be introduced in which the moving distance within the monitoring range is S1 and the length of the baseline band is S. As shown in FIG. 10, when ValidDis falls below a preset threshold (FixDis), the calculation of the vehicle speed in this baseline band is stopped, and only when ValidDis is above the threshold, the traffic state Is calculated. In other words, if a vehicle enters the baseline band, travels a short distance and exits from the baseline band, it cannot be said that the vehicle was traveling in this baseline band, so it is not considered for verification. Not. In this case, the speed calculation means 204 does not calculate the speed in the baseline band of the vehicle, and uses the vehicle only for calculation of road traffic information. Only when ValidDis is greater than or equal to the threshold value, the vehicle is considered to be traveling in the baseline band and is used as a verification target, so the travel speed of the vehicle at the distance S1 can be acquired as traffic information. .

  Furthermore, in the above-described embodiment, since the monitoring start line and the monitoring end line are set in advance by the plurality of baseline band pre-processing means 201, the baseline band as the monitoring range is determined in advance, and the period of the monitoring process Fixed throughout. Thus, in the feature matching process, all monitored objects in the baseline band are processed and compared (in other words, all vehicles in the baseline band having a fixed length are binarized and feature pixel points. It is sampled and compared to all monitored contour information and feature pixel point information in the previous image). This is not always necessary. This is because the field of view field of the camera can extend to several hundred meters. During the time period chosen as the time interval for performing the verification (eg every 2 frames, every 4 frames, every 1 second, etc., which interval is appropriate depends on the actual situation) Sometimes it only goes a few meters. In such a case, if the baseline bandwidth is long and fixed, the amount of system computation increases unnecessarily and the real-time performance of the system decreases.

  This process is further optimized if the monitoring range (baseline band range) is adaptively adjusted based on road traffic information at a previous time. FIG. 11 shows a case corresponding to this. Here, it is assumed that the monitoring system has completed the determination that the vehicle has completely entered the baseline band to be monitored by the processing of the image change portion identification unit 502 so far. In the next monitoring period (i.e., the next monitoring time or a subsequent monitoring time), the monitoring system captures an image of the entire baseline band again, and continuously executes the operations of the binarization process and the matching process. This is not always necessary. During the next period, the monitoring range (ie, the length of the baseline band from which the image is captured) can be automatically adjusted or moved depending on factors such as average vehicle speed, road speed limitations, etc. It is. In other words, it is not necessary to use the entire baseline band as a monitoring range at the next point in time for a specific vehicle. This is because the vehicle is moving and the moving range can be determined in advance based on the previously calculated average speed of the baseline band, maximum speed of the lane, and the like. In general, the movement range is much narrower than the length of the monitored baseline band. Therefore, according to one embodiment, it is desirable to adjust (move) the monitoring range in which the binarization process and the characteristic pixel point sampling are performed as the vehicle moves. Such movement is called “baseline bandwidth adjustment”. However, this type of adjustment is a virtual movement of the monitoring range related to a specific vehicle, and the baseline band obtained by the multiple baseline band pre-processing means 201 is not actually adjusted.

  For example, the baseline band of a vehicle to be monitored can be set from the rear of the vehicle to a point where the distance when moving at the maximum speed of the vehicle is added. This maximum speed can be set in advance or can be adaptively changed as required. For example, if it is known from the monitoring history that the vehicle speed does not exceed 60 km / hour on a road in an urban area, the maximum speed of the road is set to 60 km / hour, and the length of the baseline band is accordingly set. Can be set. Alternatively, the baseline band may be from the rear of the vehicle to a point where the travel distance calculated at the speed at the previous time is added. Furthermore, various other settings can be made such as setting the baseline band from the rear of the vehicle to a point obtained by adding the travel distances calculated at the previously calculated average road speed. The above-described method for setting the length of the baseline band is merely an example of the present invention, and the present invention is not limited to this. That is, the baseline band as the monitoring range does not necessarily have to be a point obtained by adding a certain distance from the rear part of the vehicle. For example, if there is no need to match the vehicle model and color and only detailed matching (eg, front vehicle license plate information) is required, the baseline band can be any distance from the center of the vehicle body. It can be up to the added point. It is also clear that any of the above-described baseline bandwidth adjustment methods can be used to reduce the length of the baseline bandwidth for performing binarization and feature pixel point extraction. With this configuration, the amount of calculation for image processing is reduced, and the real-time performance of the monitoring system is improved.

  Since the adaptive change (adjustment) of the baseline band in the present invention is not limited to any of the above, the adjustment reduces the calculation amount of the subsequent image processing, and consequently the calculation amount of the monitoring system during the monitoring process. As long as the baseline band is adjusted, any method can be used.

  As described above, immediately after a certain monitoring target (vehicle) enters the baseline band, the monitoring range of the vehicle is only a specific range near the start line of the monitoring baseline band. Absent. Therefore, when using the above settings, each of the plurality of images that the video surveillance system performs comparison (collation) and feature point extraction during successive periods is actually a part of the baseline band. It contains only content, not image content for the entire baseline band area. Obviously, this greatly reduces the amount of calculation related to the monitoring target. Needless to say, comparison (matching) and feature point extraction can be performed on the image content in the entire area of the baseline band.

  As described above, in this configuration, the baseline band as the monitoring range also moves while the vehicle moves, and its length is automatically adjusted. Therefore, it is possible to significantly reduce the length (size) of the baseline band and reduce the calculation amount of the video surveillance system.

  Although the present invention has been described above by taking the video surveillance system 200 for monitoring traffic information on the lane as an example of the video surveillance system, the present invention is not limited to this. The present invention can be applied to any video surveillance field, such as a surveillance system that monitors the movement of an article and a surveillance system that sorts articles on a conveyor belt. In this case, the monitoring target is an article on the conveyor belt, and the monitoring system divides the image of the monitoring field into one or more baseline bands in the same manner as described above, and a moving object (in this case, the belt) Perform feature extraction on the article on the conveyor to obtain the binarized outline and feature pixel points of the article (where the feature pixel points are any pixel points that can distinguish different articles) For example, in an embodiment of a baggage transfer system used in a small airfield, a pixel point obtained by sampling a baggage tag may be used as a feature pixel point). In this case, the speed calculation means calculates the speed of the article on the conveyor belt as the monitoring target in accordance with the binarized contour and the characteristic pixel point matching result in the plurality of images, and acquires the traffic information of the article on the conveyor belt. To do.

  Furthermore, the function of each means of the video surveillance system 200 described above can also be implemented by a calculation device (not shown) such as a CPU (Central Processing Unit). In this case, the computer reads the control program, interprets the control program, and executes it. A program module for executing the function of each means is recorded in the control program. This control program may be provided in the form of a recording medium, or may be downloaded to a computing device and executed by wired or wireless communication via a network. The computing device may be a general personal computer equipped with a CPU or a dedicated computing device.

  The present invention has been described above with reference to preferred embodiments. However, the present invention is not necessarily limited to the above embodiments, and various modifications can be made within the scope of the technical idea. it can.

  Further, a part or all of the above-described embodiment can be described as in the following supplementary notes, but is not limited thereto.

(Appendix 1)
Baseline band pre-processing means configured to pre-process an image of the monitored field of view captured by the imaging device, thereby obtaining one or more baseline bands;
A configuration is made to acquire a binarized contour and a feature pixel point of a moving object existing in any one or more of the baseline bands by performing feature point extraction on the moving objects in the baseline band Feature point extraction means,
A moving object identifying means configured to identify and detect the moving object by performing a matching of the binarized contours and feature pixel points of the moving object within the baseline band in the plurality of images. Baseline band video surveillance system.

(Appendix 2)
The baseline according to claim 1, further comprising speed calculation means configured to calculate a speed of the moving object for the moving object identified and detected in the plurality of images by the moving object identification means. Bandwidth video surveillance system.

(Appendix 3)
The baseline band video surveillance system of claim 2, further comprising an analysis unit configured to calculate a baseline band traffic condition based on the speed of the moving object.

(Appendix 4)
The feature point extraction means further includes:
Binarization means configured to perform binarization processing on the first color image of the field-of-view field to be monitored and extract contour information indicating the contour of the moving object;
A feature pixel point configured to perform a feature point sampling process on a pixel point of a moving object in a region located in the baseline band of the first color image, and to obtain identification information that can uniquely identify the moving object Sampling means,
The baseline band video surveillance system according to claim 1, wherein the region is identified by a contour indicated by contour information.

(Appendix 5)
The feature point extraction means further includes:
Comprising an image change portion identification means configured to identify a change portion in the current image compared to a previous image;
The binarization means and the characteristic pixel point sampling means are:
The baseline band video surveillance system according to claim 4, wherein the corresponding operation is executed only when the image change portion identification means identifies the presence of the change portion in the current image.

(Appendix 6)
The characteristic pixel point sampling means includes:
The baseline band video surveillance system according to appendix 4, wherein the region where the identification information is located is configured to sample at a higher density than other regions.

(Appendix 7)
The baseline band video surveillance system according to appendix 4, wherein the moving object is a vehicle, and the identification information is collective information about sampling pixel points of a vehicle license plate.

(Appendix 8)
The speed calculation means includes
A trajectory drawing means configured to calculate a trajectory of a moving object based on matching between a plurality of images;
The distance / speed calculating means configured to calculate a moving distance of the moving object according to the trajectory of the moving object and to calculate the speed of the moving object according to the moving distance and the moving time. Baseline bandwidth video surveillance system.

(Appendix 9)
The moving object is a vehicle;
The baseline band video surveillance system according to claim 1, wherein the matching includes at least one of contour matching, color matching, and detailed matching.

(Appendix 10)
The baseline band has a monitoring start line and a monitoring end line,
The baseline bandwidth video monitoring system according to appendix 1, wherein the monitoring start line and the monitoring end line are movable.

(Appendix 11)
The distance / speed calculating means includes:
The baseline band video surveillance system according to appendix 8, wherein the moving object velocity is not calculated when the continuous moving distance of the moving object within the baseline band is less than a certain threshold.

(Appendix 12)
The plurality of images are captured by the imaging device at a certain time interval,
The baseline band video surveillance system according to claim 1, wherein a time interval at which the plurality of images are captured by the imaging device is fixed or variable.

(Appendix 13)
Each image that is necessary for matching for a certain period and for which feature point extraction is performed is an image of either the entire region in the baseline band or a part of the entire region in the baseline band. The baseline band video surveillance system according to appendix 1.

(Appendix 14)
Pre-processing an image of the monitored field of view captured by the imaging device, thereby obtaining one or more baseline bands;
Obtaining a binarized contour and feature pixel points of the moving object by performing feature point extraction on the moving object in the baseline band for any of the baseline bands;
A baseline band video comprising: identifying and detecting a moving object by performing a match between a binarized contour of the moving object within the baseline band and a plurality of images of feature pixel points. Monitoring method.

(Appendix 15)
The baseline band video monitoring method according to claim 14, further comprising the step of calculating a velocity of the moving object identified and detected in a plurality of images.

(Appendix 16)
The baseline band video monitoring method according to claim 15, further comprising a step of calculating a traffic state of a baseline band based on the speed of the moving object.

(Appendix 17)
The step of performing feature point extraction further comprises:
Performing binarization on the first color image of the field of view field to be monitored and extracting contour information indicating the contour of the moving object;
Performing feature point sampling processing on pixel points of the moving object in a region located in the baseline band of the first color image, and obtaining identification information that can uniquely identify the moving object,
The baseline band video monitoring method according to claim 14, wherein the region is identified by a contour indicated by contour information.

(Appendix 18)
The step of performing feature point extraction further comprises:
Identifying the change in the current image compared to the previous image,
The step of performing binarization processing and the step of performing sampling processing are:
The baseline band video monitoring method according to claim 17, wherein the corresponding operation is executed only when the presence of the changed portion in the current image is identified.

(Appendix 19)
The step of performing the sampling process includes:
18. The baseline band video monitoring method according to appendix 17, wherein a region where the identification information is located is sampled at a higher density than other regions.

(Appendix 20)
The moving object is a vehicle;
The baseline band video monitoring method according to claim 17, wherein the identification information is set information on sampling pixel points of a vehicle license plate.

(Appendix 21)
Calculating the velocity of the moving object;
Calculating a trajectory of a moving object based on matching between multiple images;
The baseline band video monitoring method according to claim 15, further comprising: calculating a moving distance of the moving object according to the trajectory of the moving object, and calculating a speed of the moving object according to the moving distance and the moving time.

(Appendix 22)
The moving object is a vehicle;
The baseline band video monitoring method according to appendix 14, wherein the matching includes at least one of contour matching, color matching, and detailed matching.

(Appendix 23)
The baseline band has a monitoring start line and a monitoring end line,
15. The baseline bandwidth video monitoring method according to appendix 14, wherein the monitoring start line and the monitoring end line are movable.

(Appendix 24)
Calculating the velocity of the moving object comprises:
The baseline band video monitoring method according to appendix 21, wherein the moving object velocity is not calculated when the continuous moving distance of the moving object within the baseline band is less than a certain threshold.

(Appendix 25)
The plurality of images are captured by the imaging device at a certain time interval,
15. The baseline band video monitoring method according to appendix 14, wherein a time interval at which the plurality of images are captured by the imaging device is fixed or variable.

(Appendix 26)
Each image that is necessary for matching for a certain period and for which feature point extraction is performed is an image of either the entire region in the baseline band or a part of the entire region in the baseline band. The baseline bandwidth video monitoring method according to claim 14, wherein

201: Multiple baseline band preprocessing means 202: Feature point extraction means 203: Vehicle identification means 204: Speed calculation means 205: Analysis means 502: Image change portion identification means 504: Binarization means 506: Feature pixel point sampling means 510 : Trajectory drawing means 512: Vehicle speed calculating means

Claims (10)

Baseline band pre-processing means configured to pre-process an image of the monitored field of view captured by the imaging device, thereby obtaining one or more baseline bands;
A configuration is made to acquire a binarized contour and a feature pixel point of a moving object existing in any one or more of the baseline bands by performing feature point extraction on the moving objects in the baseline band Feature point extraction means,
A moving object identifying means configured to identify and detect the moving object by performing a matching of the binarized contours and feature pixel points of the moving object within the baseline band in the plurality of images. Baseline band video surveillance system. The base according to claim 1, further comprising speed calculation means configured to calculate a speed of the moving object for a moving object identified and detected in a plurality of images by the moving object identification means. Line-band video surveillance system. The baseline band video surveillance system of claim 2, further comprising an analysis unit configured to calculate a traffic condition in a baseline band based on the speed of the moving object. The feature point extraction means further includes:
Binarization means configured to perform binarization processing on the first color image of the field-of-view field to be monitored and extract contour information indicating the contour of the moving object;
A feature pixel point configured to perform a feature point sampling process on a pixel point of a moving object in a region located in the baseline band of the first color image, and to obtain identification information that can uniquely identify the moving object Sampling means,
The baseline band video surveillance system according to claim 1, wherein the region is identified by a contour indicated by contour information. The feature point extraction means further includes:
Comprising an image change portion identification means configured to identify a change portion in the current image compared to a previous image;
The binarization means and the characteristic pixel point sampling means are:
5. The baseline band video surveillance system according to claim 4, wherein the corresponding operation is executed only when the image change portion identification means identifies the presence of the change portion in the current image. The characteristic pixel point sampling means includes:
The baseline band video surveillance system according to claim 4, wherein the area where the identification information is located is configured to be sampled at a higher density than other areas. The baseline band video surveillance system according to claim 4, wherein the moving object is a vehicle, and the identification information is collective information about sampling pixel points of a vehicle license plate. The speed calculation means includes
A trajectory drawing means configured to calculate a trajectory of a moving object based on matching between a plurality of images;
3. A distance / speed calculating means configured to calculate a moving distance of the moving object according to the trajectory of the moving object and to calculate the speed of the moving object according to the moving distance and the moving time. The baseline bandwidth video surveillance system described. The distance / speed calculating means includes:
9. The baseline band video surveillance system according to claim 8, wherein when the continuous moving distance of the moving object in the baseline band is less than a certain threshold, the speed of the moving object is not calculated. Pre-processing an image of the monitored field of view captured by the imaging device, thereby obtaining one or more baseline bands;
Obtaining a binarized contour and feature pixel points of the moving object by performing feature point extraction on the moving object in the baseline band for any of the baseline bands;
A baseline band video comprising: identifying and detecting a moving object by performing a match between a binarized contour of the moving object within the baseline band and a plurality of images of feature pixel points. Monitoring method.

Images