JP4946228B2 - In-vehicle pedestrian detection device - Google Patents

Description

  The present invention relates to a vehicle-mounted pedestrian detection device that detects a pedestrian mounted on a vehicle and that is in a position that prevents the vehicle from advancing, and notifies the driver of the presence of the pedestrian.

2. Description of the Related Art Conventionally, an obstacle detection device that detects an object that affects a host vehicle such as a pedestrian from an image captured by an infrared camera, and issues an alarm to the driver. By consciously storing the characteristics and position information of the object to be excluded from the alarm object in the storage means, and when the object stored in the storage means is detected again, the alarm output for the object is stopped by A device that suppresses the output of an alarm bothering the driver is known (see, for example, Patent Document 1).
JP 2004-106682 A

  However, in the configuration described in Patent Document 1 described above, the driver must operate and memorize objects to be excluded from the alarm target, and if there are many excluded objects, the driver's operation burden becomes excessive. In addition, when an unnecessary object and a necessary object appear at the same time, the driver cannot perform an instruction operation and cannot store unnecessary detection.

  Furthermore, in the above-mentioned Patent Document 1, identification is performed based on the feature and position information for unnecessary detection, but the object is captured and stored differently depending on the light of the oncoming vehicle, the light of the surrounding environment, and the like. There is a problem in that it is not consistent with the features of and may not be detected as unnecessary.

  Therefore, the present invention automatically performs unnecessary detection without requiring a driver's operation based on position information where the detection of pedestrians occurs and the frequency of occurrence thereof, and reduces unnecessary detection to improve detection performance. An object is to provide a vehicle-mounted pedestrian detection device.

In order to achieve the above object, an on-vehicle pedestrian detection device according to a first aspect of the present invention is an on-vehicle pedestrian detection device having pedestrian detection means for detecting an object around a vehicle including a pedestrian,
Pedestrian position calculation means for calculating the position of the object detected by the pedestrian detection means, storage means for storing the position of the object calculated by the pedestrian calculation means, and detection by the pedestrian detection means Determination means for determining whether or not the target object is a pedestrian,
The determination means is an object detected by the pedestrian means when the number of past object detection data stored in the storage means is a predetermined number or more at the object position calculated by the pedestrian position calculation means. Determine that the object is not a pedestrian ,
In the peripheral area of the object position calculated by the pedestrian position calculation means, when the number of past object detection data stored by the storage means is a predetermined number or more, the pedestrian detection means in the peripheral area It is characterized by increasing the detection sensitivity . As a result, it is possible to prevent a non-pedestrian stationary object such as a signboard from being erroneously detected as a pedestrian, and to prevent undetected undetected even though it is a pedestrian in a busy area. .

2nd invention is the vehicle-mounted pedestrian detection apparatus which concerns on 1st invention,
The vehicle-mounted pedestrian detection device is configured such that when the number of past object detection data stored in the storage unit is equal to or greater than the predetermined number, the pedestrian detection unit no longer detects an object. The past object detection data according to the above is erased. Thereby, it will not be bothered by the removed target object.

  According to the present invention, it is possible to provide a vehicle-mounted pedestrian detection device that does not require a setting operation for erroneous detection and removal by a driver and that has few pedestrian erroneous detections.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  FIG. 1 is an example of a functional block diagram of an in-vehicle pedestrian detection device according to the present embodiment.

  The on-vehicle pedestrian detection device according to the present embodiment includes a pedestrian detection means 10, a pedestrian position calculation means 20, a storage means 30, a determination means 40, and a notification means 50. Furthermore, the pedestrian detection unit 10 may include a detection sensitivity adjustment unit 11, an imaging unit 12 such as an in-vehicle camera, or a radar 13 as necessary. Moreover, the pedestrian position calculation means 20 may also include communication information acquisition means 21 for acquiring communication information such as GPS (Global Positioning System) as necessary.

  Next, the function of each component and the outline of its operation will be described.

  The pedestrian detection means 10 is a means for detecting the presence of a pedestrian that hinders the progress of the vehicle. For example, if the vehicle is traveling forward and there is a pedestrian in front, the presence of the pedestrian is detected. The presence of a pedestrian may be detected by an image signal using an imaging unit 12 such as an in-vehicle camera, or the pedestrian is irradiated with light such as radar 13 or sound waves, and the reflected signal is used for the pedestrian. It may be configured to detect the presence of. When the radar 13 is used, for example, laser light may be used, millimeter waves may be used, microwaves may be used, and appropriate ones may be used depending on the application.

  Moreover, when detecting a pedestrian by an image signal, you may detect by various methods, For example, a pattern matching system may be used. For example, if an image pattern of a predetermined pedestrian is recognized and stored in advance and an image of the object is acquired by the imaging means 12, the image is compared with the pedestrian image stored and stored in advance. If the image patterns match, the object may be detected as a pedestrian.

  The detection sensitivity adjustment means 11 is a means for adjusting the sensitivity of pedestrian detection. If the detection sensitivity is increased, more pedestrians will be detected, but instead the number of detected objects other than pedestrians increases, so there is a high possibility that the noise for detecting objects other than pedestrians as pedestrians will increase. Become. On the other hand, if the detection sensitivity is lowered, the noise is reduced and the probability of detecting an object that is not a pedestrian is lowered, but there is a possibility that a detection failure of a pedestrian will occur instead. As described above, since the detection sensitivity is preferably adjusted to an appropriate sensitivity, the detection sensitivity adjustment means 11 may be provided as necessary. For example, when the detection sensitivity adjustment unit 11 is configured by the above-described pattern matching method, the detection sensitivity adjustment unit 11 can perform adjustment by increasing or decreasing the matching threshold value. If the threshold value is lowered, detection is performed if the image acquired by the imaging means 12 and the pattern image for comparison reference roughly match, and if the threshold value is lowered, the pattern is not matched to the details. It will not be recognized as a pedestrian.

  Information on the presence of a pedestrian detected by the pedestrian detection means 10 is sent to the pedestrian position calculation means 20.

  Note that the pedestrian detection means 10 may calculate not only the presence of the pedestrian but also the position of the pedestrian. For example, when the presence of a pedestrian is detected from an image signal, the position of the pedestrian can be detected from the direction of the camera, the tilt angle, and the like. In this case, not only the pedestrian presence information is sent to the pedestrian position calculating means 20, but also the pedestrian position information may be sent to the determination unit 40.

  The pedestrian position calculation unit 20 calculates the position of the pedestrian based on the pedestrian detection information from the pedestrian detection unit 10. The position of the pedestrian is calculated based on both the position of the host vehicle by, for example, GPS, and information indicating the positional relationship with the pedestrian such as the direction of the vehicle and the distance from the pedestrian. be able to. In order to grasp the position of the vehicle using communication information such as GPS, communication information acquisition means 21 such as an antenna or a receiver may be used. On the other hand, the positional relationship of the host vehicle with the pedestrian is determined, for example, by determining the traveling direction of the vehicle from the steering angle detected by the steering sensor and the relationship between the vehicle speed detected by the vehicle speedometer and the host vehicle position. The distance may be obtained.

  Thus, the pedestrian position calculation means 20 can calculate the position of the pedestrian by grasping the geographical position of the own vehicle and the relative position with the pedestrian based on the position of the own vehicle. it can. Information on the pedestrian position calculated by the pedestrian position calculation means 20 is sent to the storage means 30 and / or the determination means 40.

  The storage unit 30 stores information on the pedestrian position sent directly from the pedestrian position calculation unit 20 or indirectly through the determination unit 40. The information regarding the pedestrian position includes at least information capable of specifying the position of the pedestrian, and may be, for example, latitude and longitude information or XY coordinates in a GPS image. The pedestrian position information stored in the storage unit 30 uses the detection frequency at each position as data. Therefore, the pedestrian position information may be classified for each position in the storage unit 30 or converted into information indicating the number of recordings for each position. . The storage unit 30 accumulates and stores the pedestrian position information data as described above and plays a role of providing data for determination in the determination unit 40 when the data is referred to by the determination unit 40 as necessary.

  The determination unit 40 determines whether the information regarding the pedestrian position sent from the pedestrian position calculation unit 20 is correct or not, and executes necessary processing. In determining, can the past data stored in the storage unit 25 be referred to and the information regarding the pedestrian position sent from the pedestrian position calculation unit 20 based on the past data be determined as a pedestrian as it is? Or, a determination is made as to whether a non-pedestrian has been erroneously detected as a pedestrian, or there may be no detection of a pedestrian. Although a specific determination method will be described later, necessary processing is executed according to the determination result. For example, when it is determined that the sent pedestrian information is correct, the information is sent to the notification means 50 to notify the driver of the presence of the pedestrian. At that time, data may be written into the storage means 25 as new pedestrian information. Further, when it is determined that the information of the pedestrian that has been sent is erroneous detection, the notification means 50 performs control so as not to notify the presence of the pedestrian. For example, the signal to the notification unit 50 may not be output, or an error detection signal may be output and the notification unit 50 may be instructed not to output the signal. Furthermore, when it is determined from the information on the pedestrian that has been sent that there may be no detection, control such as increasing the detection sensitivity of the pedestrian detection unit 10 may be performed.

  In the determination, the pedestrian position detected by the pedestrian detection means 10 may also be collated. By comparing with the pedestrian position calculated by the pedestrian position calculating means 20, the position information of the pedestrian can be acquired more accurately and with high reliability.

  Note that the determination unit 40 may be configured by a calculation unit such as a computer. For example, the determination unit 40 may be incorporated in an ECU (Electrical Control Unit) or may be provided alone. Further, it may be configured by an FPGA (Field Program Gate Array) or the like that reads an electric circuit or a program.

  The notification means 50 notifies the driver of the presence of a pedestrian based on the command from the determination means 40. Specifically, for example, it may be executed by voice notification such as an alarm or announcement, or visual notification such as a display or a light emitting diode, or may be used in combination.

  FIG. 2 is a diagram for explaining a basic concept of determination performed by the determination unit 40 in the vehicle-mounted pedestrian detection apparatus according to the present embodiment.

  FIG. 2A is a diagram showing a relationship between a traveling vehicle, a signboard that is a non-pedestrian, and a pedestrian. In a pedestrian detection device mounted on a vehicle, for example, when detecting a pedestrian using image recognition, the image recognition is performed by distinguishing a pedestrian and a non-pedestrian from an image acquired by an imaging unit such as a camera, It is required to accurately detect only pedestrians. However, for example, a signboard on which a person image is drawn is almost the same size as a person, and the shape is similar when viewed on a two-dimensional plane. It is extremely difficult. In addition, as shown in FIG. 2 (b), even if a person is not drawn, a signboard that is slender and has the same width as a person and the upper end of the person becomes large is recognized by image recognition. It is extremely difficult to detect an object other than the object and detect unnecessary. Similarly, as shown in FIG. 2 (c), it is difficult to detect unnecessary fences or the like where the tree is a person.

  Therefore, in the on-vehicle pedestrian detection according to the present embodiment, paying attention to the fact that the pedestrian is a moving object and the non-pedestrian is a permanent stationary object, distinguishing this difference and removing unnecessary detection It was. Specifically, the position data of the object detected as a pedestrian is stored and accumulated in the storage unit 25, and every time a pedestrian is newly detected, the past detection history data at the position of the pedestrian is referred to. Like that. If past detection histories are detected a predetermined number of times or more at substantially the same position, the object can be determined not as a pedestrian but as a stationary object such as a permanent signboard. By excluding this detection target from those detected as pedestrians, it is possible to reduce the number of unnecessary detections or false detections for detecting non-pedestrians as pedestrians.

  Conversely, in places where many people come and go, such as pedestrian crossings, people may overlap or there may be people who cross their hands, etc., and the detected pedestrian image is stored for pattern reference. There may be cases where the image looks different from the image. Humans can change to any shape depending on their posture, posture, and movement, but pattern recognition does not necessarily correspond to shape patterns created by all humans. For example, if a person is crouching, It becomes very difficult to recognize that In addition, in places where people gather, such as pedestrian crossings, people may overlap each other. Therefore, in a place where many people gather, the place is recognized as a frequent area, and when a pedestrian is detected in that area, the detection sensitivity is increased, and if the image is close to a certain level, it is recognized as a pedestrian. A control method can be taken. Thereby, undetected pedestrians can be reduced.

  Next, a specific example when the vehicle-mounted pedestrian detection device according to this example is actually mounted on a vehicle will be described.

  FIG. 3 is a diagram illustrating an example of a system configuration of the on-vehicle pedestrian detection device according to the present embodiment mounted on a vehicle. In FIG. 3, the constituent elements corresponding to the constituent elements described in FIG. 1 are denoted by the same reference numerals as those in FIG. Like that.

  In FIG. 3, the vehicle-mounted pedestrian detection device according to the present embodiment refers to and records a near-infrared camera 12a, a pedestrian detection ECU 10a, a navigation 20a including a memory 30a, and a memory, and performs false detection / area determination. It is comprised from determination ECU40a which performs and alerting device 50a.

  The near-infrared camera 12a detects an image of an object that is a candidate for a pedestrian. The near-infrared camera 12a is suitable as a pedestrian detection unit because it can capture and image a thin object that is not easily scattered and can also capture images at night. A far-infrared camera may be used instead of the near-infrared camera 12a. Since they are the same infrared camera, they have similar properties. Furthermore, if possible, a camera using visible light can be used, and if it is not necessary to acquire an image, a radar or the like can be used. The image signal acquired by capturing with the near-infrared camera 12a is sent to the pedestrian detection ECU 10a.

  The pedestrian detection ECU 10a performs image recognition as to whether or not the object image acquired by the near infrared camera 12a is a pedestrian. The pedestrian detection ECU 10a stores a pedestrian image pattern as a reference image in advance, and recognizes whether or not a pedestrian is included in the object image sent from the near-infrared camera 12a by pattern matching. I will do it. The reference image may be stored in a storage unit such as a memory provided in the pedestrian detection ECU 10a, and the pattern matching calculation process may be processed by a calculation processing function in the ECU 10a. Moreover, when the pedestrian detection ECU 10a recognizes the presence of a pedestrian, the position coordinate may be calculated. Since the distance to the pedestrian can be estimated from the direction and inclination of the near-infrared camera 12a, the position coordinates of the pedestrian may be calculated and calculated in the pedestrian detection ECU 10a based on the distance. When the presence of a pedestrian is detected by the pedestrian detection means 10a, the pedestrian information is sent to the determination ECU 40a. If there is pedestrian position information, the information may also be sent to the determination ECU 40a.

  The navigation 20a has a function of knowing the position of the host vehicle using information communication means such as GPS, a function of calculating the pedestrian position based on the position of the host vehicle, and a function of storing the calculated pedestrian position in the memory 30a. Have both. These functions may be incorporated in a so-called navigation system. The navigation 20a can know the location of the vehicle in the geographic area using GPS or the like.

  In addition, the navigation 20a estimates and calculates the position of the pedestrian based on the position of the host vehicle using, for example, data that determines the positional relationship such as the direction and distance of the pedestrian such as the steering angle, the tilt angle, and the speed of the steering. To do. For example, the steering angle of the steering may be detected by a steering sensor, the tilt angle may be detected by a level meter, and the speed may be detected by a speedometer. Further, a distance sensor such as a radar may be separately provided for the distance to the pedestrian.

  Furthermore, the navigation 20a stores the pedestrian position calculated by the estimation calculation using the internal memory 30a. The pedestrian position may be stored, for example, as a geographical absolute value of latitude or longitude, may be stored as coordinates on the display screen of the navigation 20a, or both may be stored. Any coordinate system that can express the pedestrian position on the same basis may be used. Moreover, you may make it memorize | store not only a pedestrian position but the own vehicle position when a pedestrian position is detected. This is because if the vehicle position at the time of detecting the pedestrian position is also considered, more accurate determination may be expected in the determination ECU 40a.

  The pedestrian detection position information calculated in the navigation 20a is sent to the determination ECU 40a.

  The determination ECU 40a determines whether the pedestrian detection position information sent from the navigation 20a is a false detection or is in a frequent appearance area with many people. Determination ECU40a collates with the data memorize | stored in the memory 30a in the navigation 20a about the pedestrian position sent from the navigation 20a.

  In the data collation, for example, a search is made as to whether or not the data at the same coordinates as the pedestrian position sent from the navigation 20a is in the memory 30a. If there is data, pick up the data and count the number of data. If the number of data is greater than or equal to the predetermined number, it can be determined that the presence information of the pedestrian sent from the navigation 20a has erroneously detected a stationary object as a pedestrian. The determination of whether or not the coordinates are approximately the same may be set so that the coordinates determined to be the same are given a width that is approximately the same, and the coordinates are determined to be approximately the same if within the width.

  Next, for example, the determination ECU 40a determines whether the pedestrian position sent from the navigation 20a is in the frequent traffic area or not by searching the memory 30a. Since it is possible to determine whether or not the surrounding area is a frequent area depending on how many pieces of detection history data exist in the surrounding area of a predetermined area including the pedestrian position, such a surrounding area and the reference number of data are set in advance. It may be determined and whether or not it is a frequent area is determined based on the standard. Further, if the area area and the number of data are determined, the number of data existing within a predetermined area, that is, the data density can be calculated. Therefore, the frequent area may be specified based on the data density. When it is determined that the pedestrian position sent from the navigation 20a is within the frequent appearance area provided by such a reference, for example, control may be performed to increase the detection sensitivity of the pedestrian detection ECU. In a frequent appearance area, there is a high possibility that pedestrians that should be detected due to people overlapping each other are not detected. Therefore, in order to prevent such a situation, for example, pattern matching performed by the pedestrian detection ECU 10a is performed. The detection sensitivity can be increased by lowering the matching threshold level.

  The functions of the determination ECU 40a may be executed by a calculation function as a computer.

  Further, in the determination for preventing erroneous detection, the determination ECU 40a may remove past data relating to the stationary object when the stationary object that is the target of erroneous detection is removed. Since the storage area of the memory 30a is finite, the purpose is to delete unnecessary data and free the memory area. For example, when there is a plurality of detection histories at substantially the same position and there is no access to data collation for a predetermined period, it is determined that the stationary object has been removed and the data is deleted. You may set to do.

  When the determination ECU 40a determines that the pedestrian presence information sent from the navigation 20a is finally correct information, the determination ECU 40a may store the position information regarding the pedestrian presence information in the memory 30a. This is because it is added as new pedestrian information data.

  If the determination ECU 40a determines that the pedestrian presence information sent from the navigation 20a is correct information, the determination ECU 40a controls the attention drawing device 50a to perform a warning activation operation. On the other hand, when the determination ECU 40a determines that the pedestrian presence information sent from the navigation 20a is a false detection, the determination device 40a performs control so that the warning device 50a does not perform a warning activation operation.

  The alerting device 50a notifies the driver of the presence of a pedestrian by outputting a sound for alerting the driver or outputting a display for alerting the driver in accordance with a command from the determination ECU 40a.

  FIG. 4 shows a process flowchart when the vehicle-mounted pedestrian detection apparatus according to the present embodiment described in FIG. 3 is operated.

  In step 100, the near-infrared camera 12a is used to image an object that is a candidate for a pedestrian. When the image is taken, the process proceeds to step 110.

  In step 110, the pedestrian detection ECU 10a detects a pedestrian from the image of the object sent from the near infrared camera 12a. As described in FIG. 3, a pedestrian may be detected by pattern matching using an image recognition technique. If there is another method, it may be used. In step 110, pedestrian presence information and, if necessary, pedestrian location information are also detected.

  In step 120, it is determined whether or not a pedestrian has been detected. When it is determined that no pedestrian is detected, there is no need to alert the driver in particular, so the process returns to step 100 and starts again from the camera imaging. On the other hand, if it is determined that a pedestrian has been detected, the process proceeds to step 130.

  In step 130, the position information of the host vehicle is acquired by the navigation 20a using communication information such as GPS. Further, the position information of the detected pedestrian is acquired with reference to the position of the own vehicle. As described above, the position information of the pedestrian is based on the detected values such as the steering angle, the vehicle inclination angle, the vehicle speed, the distance between the vehicle and the pedestrian, and the direction and distance between the own vehicle and the pedestrian. The positional relationship may be calculated. In step 130, the past data at the position of the pedestrian is acquired from the memory 30a based on the position information of the pedestrian. If the past data is acquired, the process proceeds to step 140.

  In step 140, the determination ECU 40a calculates the detection frequency at the pedestrian detection position, and performs a determination process based on this. Specific processing is performed in step 150.

  In step 150, in the ECU determination process, first, it is determined whether or not there are N target data in the past data at substantially the same position as the pedestrian detection position. This is a step performed to determine whether or not the object detected by detecting it as a pedestrian is really a pedestrian. That is, even if a certain object is detected as a pedestrian, if there are N or more past data at substantially the same detection position, the detection object is a stationary object such as a signboard or fence that has been permanently installed at the same position from the past. This is because there is an extremely high possibility that it can be determined as a false detection. Here, N means a threshold of an arbitrary natural number, and in the past detection history, it is set to a desired value in consideration of how many pieces of data of the same position target are appropriate to be regarded as a stationary object. Can do. If there is any tendency for each region where the vehicle travels, it may be set individually according to it, or it may be set uniformly if there is no special circumstance. When there are N or more past data, the process proceeds to Step 160, and when there are not N or more past data, the process proceeds to Step 170.

  In step 160, the detected pedestrian is determined as a permanent stationary object that is a non-pedestrian, and a detection invalidation process or a process of not outputting the detection is performed. Then, the process returns to step 100 and starts again from the camera imaging.

  In step 170, it is determined whether or not there are N past data in a predetermined peripheral area including the pedestrian detection position. This is because, in areas such as pedestrian crossings where people frequently appear, the shape of the image changes due to overlapping of people, the shape of the image changes greatly due to various movements of people, and the image is recognized by walking rapidly. For various reasons such as not catching up, the possibility of undetected pedestrians that cannot be detected becomes rather higher. Accordingly, in such a frequent appearance area, a response different from the response performed in step 160 is required. Therefore, it is determined in step 170 whether or not the pedestrian detection position is in such a frequent appearance area. The determination as to whether or not it is in the frequent appearance area is performed by, for example, predetermining a circular area having a predetermined radius centered on the detection position or a rectangular area having the detection position as an intersection of diagonal lines, and data of coordinates in the area It may be determined by whether or not there are N. If the area of the area and the reference number N of the number of data in the area are determined, it can be expressed by the density of the number of data. Therefore, it may be expressed by the data number density that it is in the frequent area. N, like step 150, means an arbitrary natural number threshold. When there are N or more past data in the predetermined area, the process proceeds to Step 180, and when it is less than N, the process proceeds to Step 190.

  In step 180, it is determined that the position detected by the pedestrian is a pedestrian in the frequent appearance area, and the detection sensitivity of the pedestrian detection ECU 10a is increased from the viewpoint of preventing undetected pedestrians. In order to increase the detection sensitivity, for example, if the threshold value for determining pattern matching as a pattern match can be decreased, or the detection sensitivity of the near-infrared camera 12a can be increased, the detection sensitivity may be adjusted accordingly. The direction is changed from a perfect match of the shape to a setting that can be regarded as a pedestrian with a certain amount of match. When the detection sensitivity of the pedestrian detection ECU 10a or the near-infrared camera 12a is increased, the process returns to step 100 and starts again from the pedestrian camera imaging.

  In step 190, it is determined that the detection of the pedestrian is correct, and the driver is alerted by the alerting device 50a as a new detection target pedestrian. The alerting method may use sound or an image, but it is preferable to use both or at least use sound to prevent oversight. After alerting, go to step 200.

  In step 200, new data of the pedestrian who has given attention is written in the memory 30a and added as new data. This also becomes part of the accumulated storage of past data. After the data is written, the process returns to step 100 and the camera is imaged to prepare for detection of a new pedestrian.

  FIG. 5 is a functional block diagram showing an example of the on-vehicle pedestrian detection apparatus according to the present embodiment, which details the movement of data. In FIG. 5, for the constituent elements that have a corresponding relationship with the constituent elements described so far, the reference numerals in which alphabets are added to the same reference numerals are used so that the corresponding relations can be easily understood. Further, the description of the components having the same functions as those described so far will be omitted or simplified.

  In FIG. 5, the on-vehicle pedestrian detection device according to the present embodiment includes an on-vehicle camera 12b, a pedestrian detection unit 10b, a navigation 21b, a vehicle information detection unit 22, a determination unit 40b, a memory unit 30b, And an HMI (Human Machine Interface) unit 50b. Furthermore, the pedestrian detection unit 10b includes a detection sensitivity adjustment unit 11b. The determination unit 40 b includes a processing unit 41, a detection target determination unit 42, a detection position collation unit 43, a past data reference unit 44, and an imaging direction / position estimation unit 23.

  The vehicle-mounted camera 12b images a target object that is a candidate for a pedestrian and sends an image signal to the pedestrian detection unit 10b. The pedestrian detection unit 10b determines whether or not the input image signal corresponds to a pedestrian. As described above, the determination may be made by a pattern matching method based on a comparison between the pedestrian image pattern and the reference image. If there is another image recognition method, it may be used. Note that the detection sensitivity of pattern matching is adjusted by the detection sensitivity adjustment unit 11b. As described above, for example, the detection sensitivity may be adjusted by adjusting a threshold value regarded as matching or by adjusting the sensitivity of the camera.

  As a result of the pedestrian determination process such as pattern matching, when the pedestrian detection unit 10b detects a pedestrian, the detection information of the pedestrian is sent to the navigation 21b and the vehicle information detection unit 22, and the detection position of the pedestrian The coordinate information is sent to the detection position matching unit 43. The pedestrian detection signal is sent to the navigation 21b and the vehicle information detection unit 22 in the description of FIG. 1 in the same way as the pedestrian detection signal is sent from the pedestrian detection means 10 to the pedestrian position calculation means 20. This is because the position of the pedestrian is calculated from the detection signal of the pedestrian. On the other hand, the pedestrian detection unit 10b sends the pedestrian position detection coordinate data to the determination unit 40b because the detection position can also be estimated from the image information from the in-vehicle camera 12b. This is because the detection position can be calculated more accurately.

  Note that the calculation function of the pedestrian detection unit 10b described above may be realized by an arithmetic processor such as a computer, an image processing circuit, or the like, but an ECU is preferably used.

  The navigation 21b, the vehicle information detection unit 22, and the imaging direction / position estimation unit 23 play the same role as the pedestrian position calculation unit 20 described in FIG. The navigation 21b grasps the position of the own vehicle using GPS or the like, and the vehicle information detection unit 22 grasps the positional relationship such as the direction, inclination, distance, etc. of the own vehicle with respect to the pedestrian using detectors such as various sensors. Based on the vehicle position and the vehicle information, the imaging direction / position estimation unit 23 estimates the pedestrian position coordinates.

  The determination unit 40b includes various arithmetic processing units.

  First, the past data reference unit 44 refers to past data stored in the memory unit 30b to which the position is related based on the position information of the pedestrian calculated by the imaging direction / position estimation unit 23. Note that past detection data records are accumulated and stored in the memory unit 30b. Each data record includes, for example, the vehicle position coordinates and the pedestrian detection position coordinates as records.

  The detection position collation unit 43 collates the information of the pedestrian detection position coordinates sent from the pedestrian detection unit 10b with the position estimated and calculated by the imaging direction / position estimation unit 23, and the positions of both coincide with each other. If it is necessary to check whether or not any correction is necessary, a necessary correction process is performed to specify the position coordinates of the pedestrian.

  In the detection target determination unit 42, whether the detected pedestrian is erroneously detected based on the pedestrian position information specified by the detection position collation unit and the past data related to the pedestrian position referred to in the past data reference unit. It is determined whether or not the detected pedestrian is in the frequent appearance area and the possibility of being undetected is increasing.

  The processing unit 41 performs necessary processing based on the determination made by the detection target determination unit 42. For example, when the detection target determination unit 42 determines that the detection result is correct and the person is a pedestrian, the processing unit 41 outputs the result to the HMI unit and stores the pedestrian detection record in the memory unit 30b. The data item being written is written into the memory unit 30b. This is to notify the driver of the presence of the pedestrian and to reflect the record of detecting the pedestrian in the past data to be useful for future pedestrian detection.

  In addition, for example, when the detection target determination unit 42 determines that the detection result is a false detection and a non-pedestrian is detected as a pedestrian, the processing unit 41 sends a result invalid command to the HMI unit to call attention. Make sure that does not occur. This is to prevent the driver from being confused by the false detection notification. In the case of erroneous detection, since it is not necessary for the driver to be notified of the detection result in the end, a process that does not output the result may be performed by the processing unit, or data may be deleted. In the case of erroneous detection, it is not necessary to write to the memory unit 30b in principle. However, it is checked whether past data having substantially the same coordinates exists more than a predetermined number of times. You may make it write as detection. In this embodiment, a predetermined number or more of memory areas are required to determine erroneous detection. However, after the position of erroneous detection data is determined, it is stored as an invalid detection position, and thereafter, a predetermined number of past memory areas are stored. This is because if the data is erased, the memory area can be saved.

  For example, when the detection target determination unit 42 determines that the detection result indicates that the pedestrian is in the frequent appearance area, the processing unit 41 instructs the detection sensitivity adjustment unit 11b to increase the detection sensitivity of the pedestrian detection unit 10b. Is output. This is to prevent undetected pedestrians in the frequent appearance area.

  The HMI unit 50b may or may not generate an alert to the driver based on the processing command of the processing unit 41. It is the same as the previous explanation that various means such as voice and image can be used for alerting, and the explanation is omitted.

  FIG. 6 is a diagram illustrating an example of a storage data structure stored in the storage unit 30, the memory 30b, or the memory unit 30b applied to the vehicle-mounted pedestrian detection device according to the present embodiment. FIG. 6A shows raw data of the stored data, and FIG. 6B shows conversion data from the raw data. In FIG. 6A, the memory includes items of the coordinate information of the serial number, latitude and longitude of the own vehicle, the direction of the pedestrian from the own vehicle, and the in-image coordinates (X, Y) of the pedestrian position obtained therefrom. Is stored as data. Then, as shown in FIG. 6B, conversion data classified into substantially the same position from the raw data is created as necessary so that the calculation is easy.

  In FIG. The data of 001 and No. In the data of 002, the position of the own vehicle is substantially the same in both latitude and longitude, but the azimuth is greatly different, so the position of the target pedestrian is greatly different in the in-image coordinates. On the other hand, no. No. 002 and No. 002 data. In the data of 003, since the latitude and longitude indicating the position of the host vehicle are substantially the same and the direction is also substantially the same, the object can be estimated to be the same, and the in-image coordinates indicating the same are also substantially the same. . No. In the data of 004, the latitude and longitude indicating the position of the host vehicle are No. Although it is substantially the same as the data of 001 to 003, the orientation is completely different from each other, so the coordinates in the image of the object are also No. It is different from the objects 001 to 003.

  FIG. 6B shows converted data obtained by classifying the raw data of FIG. 6A for each position of the object and attaching a group name. No. In the data of 001, the group name is A, and No. of substantially the same coordinates. 002 and No. In 003, the group name is B. No. Since 004 is a new coordinate completely different from the above-mentioned data, the group name is also distinguished from C. Based on such conversion data, if the number of data is counted for substantially the same position, the detection frequency of the pedestrian position can be known from the past data. In addition, by determining the predetermined number N as a determination criterion, the determination unit 40 can determine the frequency.

  The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above-described embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the present invention. Can be added. Moreover, the embodiments described above can be combined with each other.

It is an example of the functional block diagram of the vehicle-mounted pedestrian detection apparatus which concerns on a present Example. It is a figure for demonstrating the fundamental view of the determination currently performed in the determination means 40 in the vehicle-mounted pedestrian detection apparatus which concerns on a present Example. It is a figure which shows an example of the system configuration | structure of the vehicle-mounted pedestrian detection apparatus which concerns on a present Example. The processing flowchart of the vehicle-mounted pedestrian detection apparatus in FIG. 3 is shown. It is a functional block diagram which shows an example of the motion of the data of the vehicle-mounted pedestrian detection apparatus which concerns on a present Example. It is the figure which showed an example of the memory | storage data structure of the vehicle-mounted pedestrian detection apparatus which concerns on a present Example.

Explanation of symbols

10 Pedestrian detection means 10a Pedestrian detection ECU
DESCRIPTION OF SYMBOLS 10b Pedestrian detection part 11 Detection sensitivity adjustment means 11b Detection sensitivity adjustment part 12 Imaging means 12a Near-infrared camera 12b Car-mounted camera 13 Radar 20 Pedestrian position calculation means 20a, 21b Navigation 21 Communication information acquisition means 22 Vehicle information detection part 23 Imaging direction Position estimation unit 30 storage unit 30a memory 30b memory unit 40 determination unit 40a determination ECU
40b determination unit 41 processing unit 42 detection target determination unit 43 detection position collation unit 44 past data reference unit 50 notification means 50a warning device 50b HMI unit

Claims (2)

A vehicle-mounted pedestrian detection device having pedestrian detection means for detecting objects around a vehicle including a pedestrian,
Pedestrian position calculation means for calculating the position of the object detected by the pedestrian detection means, storage means for storing the position of the object calculated by the pedestrian calculation means, and detection by the pedestrian detection means Determination means for determining whether or not the target object is a pedestrian,
The determination means is an object detected by the pedestrian means when the number of past object detection data stored in the storage means is a predetermined number or more at the object position calculated by the pedestrian position calculation means. Determine that the object is not a pedestrian ,
In the peripheral area of the object position calculated by the pedestrian position calculation means, when the number of past object detection data stored by the storage means is a predetermined number or more, the pedestrian detection means in the peripheral area A vehicle-mounted pedestrian detection device characterized by increasing the detection sensitivity of the vehicle. The vehicle-mounted pedestrian detection device is configured such that when the number of past object detection data stored in the storage unit is equal to or greater than the predetermined number, the pedestrian detection unit no longer detects an object. The in-vehicle pedestrian detection device according to claim 1, wherein the past object detection data relating to the vehicle is deleted.

Images