JP2008112349A - Moving object detection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a "moving object detection device" easily detecting that another moving object such as a vehicle is present in a blind area, with a low-priced configuration. <P>SOLUTION: This moving object detection device detecting the moving object present in the blind area of an automobile has: a photographing device photographing an automobile neighborhood area including the blind area of the automobile; a characteristic part extraction part setting a brighter portion or a darker portion than a photographic image as a characteristic part, and masking the other portion to extract the characteristic part; a background deletion part deleting a portion having a ground speed not more than a set speed inside the characteristic part with the portion as a background; and a decision part calculating an area of the characteristic part, and deciding that the characteristic part is the moving body present in the automobile neighborhood area when a ratio of the area and the precedingly calculated area is a set value or above. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a moving object detection device, and more particularly, to a moving object detection device that detects a moving object existing in a blind spot area of an automobile.

As indicated by the hatched lines in FIG. 12, a blind spot (blind spot area) BSP exists in the door mirror or side mirrors MRL and MRR of the automobile CR, and it is difficult to detect a vehicle traveling in the blind spot area.
Even though the rear radar provided at the rear of the car can cover the rear, it cannot cover the blind spot area. Therefore, it is necessary to provide another radar on the side surface, but it is expensive, and even if it is provided, a proximity region of about 0 to 1 m cannot be detected correctly by the radar. In addition, a method of identifying a vehicle traveling in the blind spot area by image recognition by providing a camera on the side is conceivable. However, the amount of calculation for image recognition increases, and an expensive camera is required and tends to be expensive. It was. Therefore, there is a demand for an apparatus that can easily detect the presence of another moving object such as a vehicle in the blind spot area with an inexpensive configuration.
There is a conventional technique that reliably detects a moving object even when the contrast is low when the vehicle is moving backward (see Patent Document 1). This prior art detects the presence / absence of a moving object by comparing the density conversion of the reference image behind the vehicle at the time of stopping and the density conversion of the comparison image behind the vehicle when the vehicle is moving backward.
JP 2000-335340 A

However, the prior art does not detect the presence of other moving objects such as vehicles in the blind spot area.
From the above, the object of the present invention is to easily detect the presence of other moving objects such as vehicles in the blind spot area with an inexpensive configuration.

The present invention relates to a photographing device for photographing a vehicle vicinity region including a blind spot region of a vehicle, a dark portion or a bright portion of a photographed image as a feature portion, and a feature portion extraction portion for extracting a feature portion by masking other portions, A background deletion unit that deletes a portion where the ground speed is less than or equal to the set speed as a background from the feature unit, calculates the area of the feature unit, and when the ratio of the area to the previously calculated area is greater than or equal to the set value, the feature unit A determination unit that determines that the object is a moving object existing in the vicinity of the automobile is provided.
In the object detection apparatus, when the automobile is stopped, the background deletion unit deletes a portion where the distance is equal to or greater than a predetermined value and the ground speed is equal to or less than a set speed as the background from the feature unit.
In the object detection apparatus, the feature extraction unit extracts a feature after performing a process of enhancing a dark part or a bright part of the captured image.
In the object detection device, the feature extraction unit divides the brightness into a dark range and a bright range based on a histogram of the captured image, and if the peak on the histogram does not exceed a predetermined set value, the feature extraction unit has a predetermined width L in the dark range. If the dark part is extracted as a characteristic part by a low-pass filter that passes only pixels having brightness within the range, and a peak on the histogram exceeds a set value and the peak exists only in the dark range, The bright part is extracted as a characteristic part by a high-pass filter that passes only pixels having brightness within a range of a predetermined width H, and in other cases, the low-pass filter pass range and the high-pass filter pass range are excluded, respectively. A feature is extracted by a filter that passes through pixels having brightness within the range. In this case, the low-pass filter has a pass characteristic of −σ _L to + σ _L on both sides centered on the average value aL if the average value of the dark portion of the histogram is aL and the standard deviation is σ _L, and the high-pass filter If the average value of the bright part of the histogram is bH and the standard deviation is σ _H , the average value bH is the center, and both sides have the passing characteristics of −σ _H to + σ _H.

According to the present invention, an area in the vicinity of an automobile including a blind spot area of an automobile is photographed, a dark portion or a bright portion is extracted as a feature portion from the photographed image, and a portion where the ground speed in the feature portion is equal to or lower than a set speed is the background. And the area of the feature is calculated, and when the ratio between the area and the previously calculated area is equal to or greater than a set value, the feature is determined to be a moving object existing in the vicinity of the automobile. Because there is no need to perform recognition processing for moving objects, and only a configuration and processing for determining whether or not a moving object exists, an inexpensive camera such as a fisheye camera may be used, and complicated processing is not required. Can be made inexpensively.
Further, according to the present invention, since the vehicle background is deleted during traveling and when the vehicle is stopped, it is possible to accurately determine the presence or absence of a moving object.
According to the present invention, the area of the feature is calculated, and when the ratio between the area and the previously calculated area is equal to or larger than a set value, the feature is determined to be a moving object existing in the vicinity of the automobile. Therefore, it is possible to accurately determine the presence or absence of a moving object.
In addition, according to the present invention, in order to determine the presence or absence of a moving object by extracting the dark part or bright part of the image as a characteristic part according to the pattern of the histogram, by extracting the dark part (the shadow part of the car) during the daytime, Further, at night, it is possible to accurately determine the presence or absence of an automobile in the blind spot area by detecting the bright part (headlight).

(A) Overall Configuration of Moving Object Detection Device FIG. 1 is an overall configuration diagram of a moving object detection device according to the present invention. The camera 1 is an inexpensive fish-eye camera equipped with a fish-eye lens, and is attached downward to the left and right side mirrors SMR of the automobile CR as shown in FIGS. The angle of the fish-eye lens is 190 ^0, 3 even downward (A), it is possible to shoot the automobile neighboring region including the blind spot of the motor vehicle as shown in (B).
The feature extraction unit 2 uses a dark part or a bright part as a feature part from a photographed image obtained from the camera and masks other parts to extract the feature part, and details will be described later with reference to FIG. The background deletion unit 3 deletes a portion where the ground speed in the extracted feature portion is equal to or lower than the set speed as the background. That is, the background deletion unit 3 sets the vehicle speed detected by the vehicle speed detection unit 4 to V,
a) If V ≠ 0, exclude the moving area with a relative speed of −V, in other words, the stationary area with a ground speed of 0 from the extracted features,
b) If V = 0, the relative speed is 0, in other words, the ground speed is 0, and a static region whose distance from the vehicle is equal to or greater than a threshold is excluded. This is because the moving object detection device of the present invention detects the presence or absence of an object (vehicle) moving in the blind spot area.
The moving object presence / absence determination unit 5 calculates the area of the feature part from which the background has been deleted, and when the ratio between the area and the previously calculated area is equal to or greater than a set value, the extracted feature part is a moving object that travels in the vicinity of the automobile. (Vehicle, motorcycle, etc.) This is because if the area ratio is equal to or greater than the set value, it can be regarded as running in parallel with the automobile. If there is a moving object in the vicinity of the automobile, the determination result output unit 6 outputs that fact by voice or image.

(B) Feature Extraction Unit FIG. 4 is a configuration example of the feature extraction unit 2.
The original image storage unit 11 stores image information (light / dark image information) captured and captured by the camera 1, and the histogram creation unit 12 creates a histogram based on lightness information included in the image information. FIG. 5A shows an example of a histogram, in which the horizontal axis is brightness, and the vertical axis is the number of pixels having the brightness (frequency). The smoothing unit 13 finely adjusts the lightness interval in the histogram where the frequency is high, and increases the lightness value in each pixel so that the frequency is substantially equal by reducing the interval in the low frequency portion. That is, the lightness is requantized so that the histogram has a uniform distribution. If the brightness of the image gathers too close to a specific value, it will be difficult to see small changes in brightness, and it will be difficult to understand the fine features of the image. Therefore, smoothing is performed to make the image easy to see with increased contrast. FIG. 5B is a histogram after smoothing.

  The tone curve adjusting unit 14 emphasizes a dark part or a bright part of an image by a preset tone curve TNC. FIG. 5C is an explanatory diagram of the tone curve TNC, which is superimposed on the smoothed histogram. The horizontal axis is the lightness of the same scale as the histogramgram, and the vertical axis is the lightness after conversion by the tone curve TNC. The brightness of each pixel is converted from the tone curve TNC. For example, the brightness of a pixel having brightness D1 is d1. The tone curve TNC in FIG. 5C emphasizes the brightness of the dark part DAR, the intermediate part MAR, and the bright part BAR. However, the tone curve TNC can be set so as to emphasize only the dark part DAR and the bright part BAR. . The image shown in FIG. 6 captured by the camera by the tone curve adjustment that emphasizes the dark part DAR and the bright part BAR is the image shown in FIG. In FIG. 7, SKY is a bright part in the sky, and CSH is a dark part in the shadow of the vehicle, and both the sky SKY and the shadow CSH of the vehicle are emphasized and appear white.

The feature part extraction filter creating unit 15 creates a low-pass filter LPF that extracts a dark part, a high-pass filter HPF that extracts a bright part, and a filter BPF that extracts an intermediate part as filters, and outputs predetermined filter characteristics.
The filter characteristic determination unit 16 determines which filter is used as a feature extraction filter, and divides the brightness of an image into a dark range DKA and a bright range BRA based on a histogram. Then, the filter characteristics decision unit 16, as shown in FIG. 8 (A), if the peak on the histogram exceeds the set value N _TH image regarded as a daytime image, the low-pass filter so as to output a dark portion LPF is determined as a feature segmentation filter. That is, the low-pass filter LPF that passes only pixels having brightness within the predetermined width L in the dark range DKA is determined as the feature extraction filter. Thereby, the dark part which is the shadow of a car can be output as a characteristic part in the daytime.
Further, as shown in FIG. 8B, the filter characteristic determination unit 16 regards the image as a night image if the peak on the histogram exceeds the set value N _TH and the peak exists only in the dark range DKA. The high pass filter HPF is determined as a feature segmentation filter so that a bright part can be output. That is, the high-pass filter HPF that passes only pixels having brightness within the predetermined width H in the bright range BRA is determined as the feature extraction filter. Thereby, a headlight can be detected at night.
In other cases, for example, as shown in FIG. 8C, if the peak exists in both the dark range DKA and the bright range BRA, the image is regarded as an evening image, and the pass range of the low-pass filter and the high-pass filter A filter FIL that passes through pixels having brightness in a range excluding each of the pass ranges is determined as a feature segmentation filter.

The feature extraction filter creation unit 15 creates the low-pass filter LPF and the high-pass filter HPF as follows. FIG. 9 shows a filter creation processing flow. First, the feature extraction filter creation unit 15 calculates the average values aL and bH of the dark part DKA and the bright part BRA (step 101). Next, standard deviations σ _L and σ _H of the dark part DKA and the bright part BRA are calculated (step 102). When the above calculation is completed, the feature extraction filter creation unit 15 creates a low-pass filter LPF centered on the average value aL and having pass characteristics on both sides −σ _L to + σ _L (width L = 2σ _L ). Further, a high-pass filter HPF having a pass characteristic centered on the average value bH and having both sides −σ _H to + σ _H (width L = 2σ _H ) is created (step 103). Note that −n · σ _L to + n · σ _L and −n · σ _H to + n · σ _H can be set as the width of the passband. n is an adjustment coefficient.
The feature cutout unit 17 masks pixels other than the passband of the filter using the filter input from the feature extraction filter creation unit 15, and extracts the feature portion through only the pixels in the passband. For example, if the peak on the histogram does not exceed the set value N _TH , the filter characteristic determination unit 16 determines the low-pass filter as the feature segmentation filter, and the feature extraction filter creation unit 15 uses the low-pass filter filter LPF as the feature segmentation unit. 17 The feature cutout unit 17 extracts a feature portion (dark portion) using the input low-pass filter LPF.
FIG. 10 is an example of an image obtained by extracting a dark part (an automobile shadow) CSH as a characteristic part from the image of FIG. 7 input from the tone curve adjustment part 14 using a low-pass filter LPF.

(C) Moving Object Detection Process Flow FIG. 11 is a moving object detection process flow of the present invention.
The feature extraction unit 2 extracts a dark part or a bright part from the image information captured by the camera as a feature part (see FIG. 10, step 201). Next, the background deletion unit 3 removes unnecessary portions such as a portion outside the camera field of view and the upper portion of the image from the extracted image (FIG. 10) (step 202), and excludes an unnecessary background from the feature portion (step 203). The background deletion unit 3 sets the vehicle speed detected by the vehicle speed detection unit 4 to V,
a) If V ≠ 0 (running), exclude the region where the relative speed is −V, in other words, the stationary region where the ground speed is 0, from the feature,
b) If V = 0 (stopped), the relative speed is 0, in other words, the ground speed is 0, and a stationary region whose distance from the automobile is equal to or greater than the threshold is excluded from the feature portion. This excludes distant trees, motorcycles, people, building windows, etc. that are visible in the feature image, and also excludes roads.
Next, the moving object presence / absence determination unit 5 calculates the area of the feature (white car part) from which the background has been deleted (step 204), calculates the ratio of the area to the previously calculated area (step 205), and It is checked whether the area ratio is equal to or greater than the set value (step 206). If the area ratio is equal to or greater than the set value, it is determined that the extracted feature is a moving object (vehicle, motorcycle, etc.) that travels in the vicinity of the automobile. The output unit 6 outputs the presence of a moving object in the vicinity of the automobile by voice or image (step 207).
As described above, according to the present invention, it is possible to detect a moving object (such as a two-wheeled vehicle, a normal vehicle, or a large vehicle) existing in a blind spot with an inexpensive configuration.

It is a block diagram of the whole moving object detection apparatus. It is camera explanatory drawing to a motor vehicle. It is explanatory drawing of the picked-up image by a fisheye lens. It is an example of composition of a feature extraction part. FIG. 6 is an explanatory diagram of a histogram and image processing (smoothing and enhancement processing) using the histogram. It is the example of an image image | photographed with the camera. It is the example of an image processed by the tone curve adjustment which emphasizes a dark part and a bright part. It is filter characteristic determination processing and filter characteristic explanatory drawing. It is a filter creation processing flow. It is an example of an image extracted as a feature part. It is a processing flow of the moving object detection of this invention. It is a blind spot (dead angle area) explanatory drawing of a motor vehicle.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Camera 2 Feature extraction part 3 Background deletion part 4 Automobile speed detection part 5 Moving object presence determination part 6 Determination result output part

Claims (5)

In a moving object detection device for detecting a moving object present in a blind spot area of an automobile,
An imaging device that captures an area near the automobile including the blind spot area of the automobile,
A feature extraction unit that extracts dark portions or bright portions of the captured image as feature portions and masks other portions to extract the feature portions;
A background deleting unit that deletes a portion of the ground speed below the set speed as a background from the feature unit;
A determination unit that calculates an area of the feature part, and determines that the feature part is a moving object existing in a vehicle vicinity region when a ratio between the area and the previously calculated area is equal to or greater than a set value;
A moving object detection device comprising: When the automobile is stopped, the background deletion unit deletes a portion having a distance of a predetermined value or more and a ground speed of a set speed or less as a background from the feature unit,
The moving object detection device according to claim 1. The feature extraction unit extracts a feature after performing a process of enhancing a dark part or a bright part of the captured image;
The moving object detection device according to claim 1. The feature extraction unit divides the brightness into a dark range and a bright range based on a histogram of the photographed image. If the peak on the histogram does not exceed a predetermined set value, the brightness within a range of a predetermined width L in the dark range is calculated. If the dark portion is extracted as a characteristic portion by a low-pass filter that passes only the pixels having the same, and the peak on the histogram exceeds the set value and the peak exists only in the dark range, it is within the range of the predetermined width H in the bright range. The bright part is extracted as a characteristic part by a high-pass filter that passes only pixels having the brightness of, and in other cases, the brightness within a range excluding the pass range of the low-pass filter and the pass range of the high-pass filter, respectively. Extracting features by a filter that passes through the pixels it has,
The moving object detection apparatus according to claim 1, wherein the moving object detection apparatus is a moving object detection apparatus. The low-pass filter has a pass characteristic of −σ _L to + σ _L on both sides of the average value aL when the average value of the dark portion of the histogram is aL and the standard deviation is σ _L, and the high-pass filter is If the average value of the bright portion of the histogram is bH and the standard deviation is σ _H , the average value bH is the center, and both sides have a passing characteristic of −σ _H to + σ _H.
The moving object detection device according to claim 3.

Images