JPS61200410A - Detection of moving body - Google Patents

Abstract

PURPOSE:To realize day and night watching, by adopting a rear image pick-up method and introducing an image at the optimum timing after checking density change inside an independent range at the lower part of the screen. CONSTITUTION:Images to be processed by an image processor 22 are divisible into 2 classifications: The No.1 image is for detecting whether a moving body 1,000 is introduced into the image pick-up field as an input of not and the No.2 is a whole image of the moving body 100. The No.2 image is caught by a timing signal obtained basing upon the No.1 image and the image of the whole image is stored in an image memory 24. For input judgment of the moving body 100 by using the No.1 image, binarization with the threshold value previously set by a binarizing circuit 23 is made and a binary frequency distribution of the output image is obtained or the preprocessed result is judged by obtaining directly the density frequency distribution by frequency distributing cumulative processor 25. The rear-image pick-up is performed and by processing further the density data inside a small area, the permanent optimum-timed images can be obtained day and night.

Description

[Detailed Description of the Invention] Industrial Application Field The present invention is a method of capturing a moving object as a two-dimensional image of an ITV TV camera, performing various image processing on the image, and detecting the number etc. of the moving object. Regarding.

, [Background of the Invention] Generally, there is an increasing need to recognize the number and shape of a moving object such as a vehicle by using an ITV TV camera image processing device without relying on humans.

Conventionally, as shown in Fig. 4, the optimal capture timing has been determined using special infrared 40, magnetic 41, ultrasonic 42, etc. sensors. Since there were practical drawbacks, in order to solve these problems, a method of determining and importing only image data was invented as described in Japanese Patent Application No. 59-98428.

However, this method involves photographing a moving object from the front or top, and when 24-hour monitoring is required, the headlights cause no-rations and are difficult to capture. There was a risk that the optimum import timing could be misjudged.

[Purpose of the invention]

The purpose of the present invention is to process images taken from the rear of a moving object such as a vehicle, thereby making it possible to capture images at night, which is a drawback of conventional front-viewing methods, and to enable high-speed recognition in real time during the day and night. An object of the present invention is to provide a method for detecting a moving object.

a. Summary of the Invention] In order to make the present invention compatible with monitoring at night, the camera is photographed from the rear, and in order to detect a moving object entering from below within the imaging sight needle, an independent nine at the lower part of the screen is used. By examining density changes within a region, images can be manually captured at the optimal timing, making daytime and nighttime monitoring possible.

[Embodiments of the Invention] Examples of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a configuration system of the present invention, and FIG. 2 is a diagram showing a hardware configuration according to the present invention.

As shown in FIG. 1, the ITv camera 10 is installed at a distance of height H, perpendicular to the route of travel. 100 is a moving object,
200 indicates an image processing device. If the viewing angle of this ITV camera 210 is 2φ, and the angle between the center point O on the traveling road surface and the traveling road surface is θ, then the following relational expression holds true.

Lo=H/sinθ L t = H/ tanθ L z = H/ tan (θ+φ) Ls = H/l
a! +(θ−φ) L4=L! Lt Ls=Ls Lt As a specific example, the angle θ between the ITv camera 10 and the traveling road surface
= 30°, ■Height of TV camera H = 5m.

D=λQm, and of course the speed of an actual vehicle is not constant.
Assuming n/h, Lo = 10.0 m.

L! =8.7 rn-s L1=5.7 ms Ls
” 14.8 m5L4 = 9.0 m, L8 =
6.0 m, φ=11.3', so the field of view in the vertical direction of the screen is L4 m, and when a vehicle passes through this distance, it takes T=L4/V=128 x 10-3 time. Also, when shooting with an ITV camera (non-interlace mode), it takes about 16ms to obtain one image. That is, when a vehicle passes by at the above speed, there is a screen movement of 1/8 before the next image is obtained.

This means that, for example, when the vehicle height is hm, eight images are obtained before the rear end of the vehicle passes within the screen.

Therefore, in addition to capturing each image, the gray level data from the TV camera is binarized, and "1" is applied to each image.
By performing a series of processes at high speed, such as calculating the frequency distribution of , it becomes possible to determine online whether a moving object has entered the imaging field of view.

An image processing apparatus as shown in FIG. 2 can be considered as one configuration for realizing such processing.

As shown in the figure, an AD converter 2011 takes in a read video signal from the ITV camera 10 and performs AD conversion. The selector 21 normally selects the AD output of the AD converter 20.

The image processing processor 22 takes in the output of the selector 21 and performs preprocessing.

This preprocessing involves noise removal and spatial differentiation. Image H to be processed by this image processing processor 22
There are 2m types. The first is an image for detecting whether the moving object 100 has entered the imaging field of view. The second is an image of the entire moving object obtained at 9 o'clock when the entire moving object 100 itself is input within the imaging field of view. This second image is captured by the timing signal obtained based on the first image, and the image of the entire screen is stored in the image memory 24.

In the method of determining the input of the moving object 100 using the first image, after performing preprocessing such as noise removal in the image processing processor 22, the binarization circuit 23 performs a binary value using a preset threshold value. and the binary frequency distribution of the output image (10”
, '''1#) and determine the input of a moving object based on that value. A method can be implemented in which the distribution is determined, the concentration average level, the concentration variance, etc. are determined, and a similar determination is made based on the values.

Next, a specific example of the process will be shown using FIGS. 3 and 4. Third
As shown in the figure, if the moving object 100 is not present in the screen and approaches from below at a speed of Nib/h, the density average level, density dispersion, etc.
The area etc. when binarized changes from the case where there is no intrusion. In FIG. 3, a 40H infrared sensor, 41 a magnetic sensor, and 42 an ultrasonic sensor. To explain this with reference to Figure 4, area 5 in (A) is an independent area at the bottom of the screen.
The concentration data of 0 changes due to the approach of the moving object. This change makes it easier to detect input from moving objects within the imaging field of view (b))
, and then C), and the same density data as in (a) is detected in the same area 50, and the moment of 97 is determined as the optimum timing for capturing the image. Here, 10
1 indicates the front part of the moving body, 102 the middle part of the moving body, and 103 the rear part of the moving body.

By using such a method, compared to forward photography,
It is fully capable of detecting moving objects such as vehicles at night, and enables more reliable detection of moving objects.

Next, a flowchart of the process will be explained using FIG.

First, an image is input using the image input subroutine, and when the density frequency distribution of the image exceeds a threshold value, it is determined to be an input from a moving object, and the moment when the subsequent density frequency distribution becomes below the threshold value is determined to be the optimal image. The image is then captured and processed later.

Next, other embodiments will be described. In order to speed up processing, it is necessary to know the optimal image capture timing early. However, as shown in Figure 6, the raster scan of current TV cameras uses the upper left of the screen as the starting point 60, and
It scans line by line and sets the lower right as the end point 61,
Therefore, in order to obtain information within a small area at the bottom of the screen,
This is about 16m5ec later than the scanning start time (non-interlaced mode). Therefore, with the hardware as shown in FIG. 9, it takes 1/60 (8) to capture an image, and to time is required to determine the input of the moving body 100 based on the image. That is, it is possible to determine the input of a moving object only once every two frames, and it may be impossible to follow a high-speed moving object. Therefore, for example, by installing the camera body upside down,
If raster scanning is performed from the moving object approach direction as shown in the figure, and processing is performed by applying a window to only the necessary parts as shown in Fig. 10, it will take 11 hours to capture the image with hardware as shown in Fig. 10. It only takes t2 time for input determination. In other words, one judgment can be made every 1/60 (8),
Judgment can be made in real time at video rate.

In this case, the image will be upside down, but in the case of recognizing characters, numbers, etc., this can be done by changing the recognition processing method in advance. (9) By storing image data in memory in advance so that it can be displayed normally even when displayed on a monitor, etc., it is possible to always process the image data normally.

〔Effect of the invention〕

According to the present invention, it has been difficult to capture images at night due to the headlights of moving objects such as vehicles, but it is now possible by capturing images from the rear and processing density data within a small area. , images can always be obtained at the optimal timing, day or night. As a result, the target area can be extracted in a shorter time than the conventional front photographing method, which has the effect of shortening the overall processing time and increasing the recognition rate of the number and shape of a moving object such as a vehicle.

[Brief explanation of drawings]

FIG. 1 is a diagram explaining the configuration system of the present invention, FIG. 2 is a diagram showing the hardware configuration according to the present invention, FIGS. 3 and 4 are diagrams showing the entering state of a moving object, and FIG. Flowcharts showing the processing of the present invention; FIGS. 6, 7, and 8 are diagrams showing an example of raster scanning for explaining other embodiments of the present invention; FIGS. 9 and 10 are diagrams showing other examples of raster scanning. In order to explain the embodiment, here is a time chart of Doware. 10... ITV camera, 100... Mobile object, 200
. . . Image processing device, 50 . . . Specific target area for processing.

Claims (1)

[Claims] 1. In an image processing device that determines whether a moving object has entered the imaging field of view based only on image data, an independent area is provided on the screen in the direction in which the moving object enters. The moment when the image data of the area changes is assumed to be the moment when the front part of the moving object enters the imaging field of view, and the image data of the area changes.
A method for detecting a moving object, characterized in that the moment when the state changes to the absence of the moving object is assumed to be the moment when the rear of the moving object is input into the imaging field of view, and an image is captured as the optimum timing. 2. In the method for detecting a moving object according to claim 1, the image is captured from a TV camera, and the image is captured from a TV camera.
A method for detecting a moving object, characterized in that the optimal image capturing timing can be detected early by inverting the camera and performing raster scanning from the direction in which the moving object approaches.