JP6348368B2 - Object detection device - Google Patents

Description

  The present invention relates to a target detection device that detects a target region where a predetermined target appears from an input image.

  Search processing by a classifier or the like is performed in order to detect a person region or the like from an input image taken by a surveillance camera or the like. Since the position and size of the target in the input image are generally unknown, in this search process, a window area is set at each position in the input image, and the image in the window area is input to a discriminator or the like. Then, a window region whose score output from the discriminator or the like exceeds a threshold is extracted as a target candidate region.

  At this time, not only a candidate area having a true target position and size is extracted for one target, but also a plurality of candidate areas having different positions and / or sizes are extracted in the vicinity or inside thereof. The Therefore, the candidate areas having overlap are grouped as a plurality of candidate areas extracted for one target, and the candidate area having the highest score for each area group is selected as the target area. Yes.

  Here, the area group may be extracted even in an area where only the background is shown (hereinafter, background area). Therefore, conventionally, the number of candidate areas extracted in duplicate is obtained, and candidate areas in which the obtained number falls below a threshold value are deleted. This uses the tendency that the number of candidate areas extracted in the background area is small.

JP 2010-160640 A

  However, even if it is a background area, a relatively large number of candidate areas may be extracted in an overlapping manner, and even an area in which an object is captured is redundantly extracted due to influences such as a change in posture of the object. The number of candidate areas may be relatively small.

  Therefore, if attention is paid only to the overlap number as in the prior art and an area group in which the overlap number falls below the threshold value is deleted, the area group including the target may be erroneously deleted. Alternatively, if the overlap threshold value is set high, there is a risk of failing to detect the region group of the target region.

  Even if a candidate region represents a true target, the score may be lower due to changes in the posture of the target, etc., so simply increase the threshold for extracting the candidate region and suppress the extraction of region groups in the background region. Then, there was a problem of failing to detect the target.

  As described above, the conventional technique has a problem that it is relatively easy to appropriately distinguish the target area group from the background area group.

  The present invention has been made in view of the above problems, and an object detection device capable of detecting a target with high accuracy from an input image while preventing erroneous detection of the target while preventing erroneous detection of the target in the background region. The purpose is to provide.

  An object detection device according to the present invention detects an object region in which a predetermined object appears in an input image, and is an index representing the likelihood that the object exists in an attention region set in the input image A storage unit that stores in advance an index value calculation function for calculating a value using a feature amount extracted at various points in the input image, and sets the attention area at a plurality of positions in the input image Then, an index value calculation unit that calculates the index value in the attention area by the index value calculation function, and the attention area that has the index value equal to or greater than a predetermined first threshold is extracted as a candidate area. In addition, a region group generation unit that generates a region group composed of a plurality of the candidate regions satisfying a predetermined overlapping relationship between the candidate regions, and belonging among the region groups An area group deletion unit that deletes the number of candidate areas that are equal to or less than a predetermined number threshold and the index value of the candidate area to which the candidate area belongs is equal to or less than a second threshold that is set higher than the first threshold; A target region determining unit that determines the target region from each of the region groups that have not been deleted by the region group deleting unit.

  In the object detection device according to the present invention, the second threshold value when the number of the candidate regions belonging to the region group is equal to or less than the number threshold value is set according to the number, and each of the second threshold values The set value is preferably set to a smaller value as the number corresponding to the set value increases.

  According to the present invention, it is possible to accurately reduce background misdetection while preventing an object from being missed.

1 is a schematic block configuration diagram of a person detection device according to an embodiment of the present invention. It is a schematic diagram which shows the example of an input image and a reduction image. It is a flowchart which shows operation | movement of the outline of the person detection apparatus which concerns on embodiment of this invention. FIG. 10 is a schematic process flow diagram of an area group generation unit. It is a typical image explaining the subsequent process with respect to the candidate area | region extracted by the area | region group production | generation part. FIG. 10 is a schematic process flow diagram of an area group deletion unit.

  Hereinafter, embodiments of the present invention (hereinafter referred to as embodiments) will be described with reference to the drawings. The target detection apparatus according to the present embodiment is a person detection apparatus 1 that targets a person shown in an image as a detection target.

[Configuration example]
FIG. 1 is a schematic block diagram of a person detection device 1 according to the embodiment. The person detection device 1 includes an image input unit 2, a control unit 3, a storage unit 4, and an output unit 5. The image input unit 2, the storage unit 4, and the output unit 5 are connected to the control unit 3.

  The image input unit 2 is, for example, an imaging device such as a surveillance camera or a recording device such as a digital video recorder that records video, and outputs an image to the control unit 3. Hereinafter, an image input from the image input unit 2 to the control unit 3 is referred to as an input image.

  The control unit 3 is configured using an arithmetic device such as a CPU (Central Processing Unit) or a DSP (Digital Signal Processor). The control unit 3 processes the input image from the image input unit 2 to determine the presence / absence of a person and outputs the determination result to the output unit 5. For this purpose, the control unit 3 reads out and executes a program from the storage unit 4, and executes an image reduction unit 30, a feature amount extraction unit 31, an index value calculation unit 32, a region group generation unit 33, a region group deletion unit 34, and a target region. It functions as the determination unit 35.

  The image reduction unit 30 reduces the input image at a plurality of preset magnifications in response to the various sizes of the person captured in the input image. This makes it possible to detect person areas of various sizes without changing the size of the window area set for detecting a person in the image. For example, the image reduction unit 30 sequentially reduces the input image at a predetermined interval until it reaches a predetermined minimum width or height, and generates a reduced image. For example, the reduction ratio is set to 10 levels until the vertical and horizontal sizes are halved. For example, the image 100 shown in FIG. 2A is an input image of the original size, and the images 110 and 120 shown in FIGS. 2B and 2C are examples of input images obtained by reducing the image 100.

  The feature amount extraction unit 31 divides each of the original size input image and the reduced input image into predetermined block sizes, and extracts the feature amount from the image of each block. Conventionally known features such as Histograms of Oriented Gradients (HOG), Local Binary Pattern (LBP) features, Haar-like features, etc. It can be used alone or in combination.

  The index value calculation unit 32 sets a window area (attention area) of a predetermined person size as a frame for detecting a person at each position in the original size input image and the reduced input image, and A score, which is a multi-valued index value representing the likelihood that the target exists in the window area, is calculated using feature values extracted at various points in the input image and an index value calculation function learned in advance. For example, the index value calculation unit 32 calculates the score for the window area by inputting the feature amount in each window area to the index value calculation function, or takes into account the posture variation that the arm part of the person protrudes from the window area Then, a feature amount within a predetermined range in and around the window area is input to the index value calculation function to calculate a score for the window area.

  In FIG. 2, an example of the rectangular window region 101 set in the images 100, 110, and 120 is indicated by a dotted line. The index value calculation unit 32 repeatedly sets the window area 101 while gradually shifting it, and scans the entire image. For example, scanning of the window region 101 starts in the horizontal direction from the upper left of the image. The horizontal scanning is repeated while shifting the vertical position little by little.

  In this embodiment, the index value calculation function is a discriminator that discriminates between “persons” to be detected and those other than “persons”. The discriminator is learned in advance using a large number of images in which “people” are reflected and a large number of images in which “people” are not reflected, and is stored in an index value calculation function storage unit 40 described later. The index value calculation unit 32 calculates a score by giving a feature amount to the classifier according to the position of the window region. The index value calculation unit 32 stores rectangular information (position, width, and height in the input image) of the window area and its score in an index value storage unit 41 described later. For example, the upper left coordinates of the rectangle forming the window area are stored as the position of the window area in the input image.

Region group generation unit 33, from the index value storage section 41, extracts the window region is first thresholds T ₁ or the score is predetermined as a candidate region, a predetermined overlapping relation on the candidate regions mutually A region group composed of a plurality of regions to be filled is generated. Specifically, the region group generation unit 33 generates region group information by assigning the same label number to candidate regions having a predetermined overlap or more. At that time, a candidate area having a high score is preferentially set as the core of the group. Details will be described later in the description of the operation. The label number of each candidate area assigned by the area group generation unit 33 is stored in the area group storage unit 42 described later together with the rectangle information and the score.

  The area group deletion unit 34 deletes the area groups stored in the area group storage unit 42 that match the deletion condition from the area group storage unit 42. Here, experimentally, an area group including a person area has a larger number of candidate areas than an area group not including a person area, and the number of candidate areas temporarily decreases due to posture variation, occlusion, and the like. Even in this case, it was found that the score of the candidate area tends to be higher than the score of the candidate area of the area group not including the person area. Therefore, the deletion condition is that the area group satisfies both of the following two requirements (R1) and (R2).

  (R1) The number of candidate regions belonging to the region group is equal to or less than a predetermined number threshold M.

(R2) The scores of all candidate regions belonging to the region group are equal to or lower than a second threshold T ₂ determined higher than the first threshold T ₁ .

  The number threshold M is determined experimentally and empirically depending on the performance of the index value calculation function, the setting of the first threshold, the setting of the reduction ratio, and the like. The number of candidate regions can be obtained for each group, and the maximum value obtained from these groups can be set as the threshold value M. That is, the number threshold M is a threshold determined based on the region group having the largest number among the groups not including the person, and therefore, the region group that may not include the person and the region group that surely includes the person. It can be estimated that the threshold value is a boundary.

First thresholds T ₁ may be set based on preliminary experiments, it is set to a lower value so as not fail to detect a true person area. For example, the first thresholds T ₁ may be the minimum value of the scores calculated for the true person region in the experimental data. Also the second threshold value T ₂ can be set based on preliminary experiments. The second threshold value T ₂ is higher than the first threshold value T ₁ , but whether or not the requirement (R 2) regarding the second threshold value T ₂ is applied by the region group deletion unit 34 is a problem for the region group. It is limited to satisfying (R1). Therefore, for example, the maximum score is obtained for each area group including the true person area and the number of candidate areas belonging to the experimental data is M or less, and the minimum value of the maximum scores obtained from these groups is set as the second threshold value. it can be a T _2.

Since the second threshold value T ₂ is a threshold determined based on the most score lower area group in the region group including a person, a threshold at the boundary between the region groups that do not contain area groups and persons including the same person Can be estimated.

  The target area determination unit 35 obtains a final person area from the area group stored in the area group storage unit 42. The target region determination unit 35 determines one person region for each region group left by the region group deletion unit 34 and stores the region information of the person region in the target region storage unit 43 together with the score. For example, the target area determination unit 35 selects one candidate area having the maximum score in each area group as the final person area. Alternatively, the target area determination unit 35 calculates a final person area by averaging the candidate areas constituting the area group for each area group. When obtaining the average, the score may be weighted.

  When at least one final person region is detected from the input image, the control unit 3 outputs the information to the output unit 5.

  The storage unit 4 is a storage device such as a ROM (Read Only Memory), a RAM (Random Access Memory), and a hard disk, and stores programs and data used by the control unit 3. The storage unit 4 inputs and outputs these programs and data to and from the control unit 3. The storage unit 4 has functions as an index value calculation function storage unit 40, an index value storage unit 41, an area group storage unit 42, and a target area storage unit 43.

The index value calculation function storage unit 40 uses a feature amount extracted by each block in the input image to obtain a score, which is an index value representing the likelihood that the target exists in the window area set in the input image. calculated for the index value calculating function for, and the first thresholds T ₁ are stored in advance. As described above, the index value calculation function is a discriminator. Specifically, the index value calculation function is obtained by applying a support vector machine (SVM) to learning images of human beings and learning images other than human beings. The parameters of the discriminator are stored in the index value calculation function storage unit 40. When linear SVM is used as the learning algorithm, the parameter of the discriminator is a weight vector generated from the learning image. This weight vector is a weight for each element of the feature amount. The weight vector is adjusted in learning so that the inner product of the weight vector and the feature amount extracted from the learning image is identified as a person when it is greater than 0, and when it is equal to or less than 0, it is identified as a person other than the person. The value of the inner product of the quantity and the weight vector represents the score. Therefore, the threshold for discriminating the score between a person and a person other than a person is 0 in principle, and the first threshold T ₁ is normally set to 0. However, as described above in the present embodiment, in order to reduce the error identified as other than human human, leave the first thresholds T ₁ is set to a value less than 0.

  As the learning algorithm of the discriminator, various conventionally known ones such as the AdaBoost method can be used in addition to the SVM.

  In addition, a pattern matching device can be used instead of the discriminator. In this case, the score is the reciprocal of the distance between the average pattern of feature values extracted from the human learning image and the feature value of the input image, and the index value. The calculation function can be a function having the score as an output value and the feature quantity of the input image as an input value.

  The index value storage unit 41 stores information on each window area calculated by the index value calculation unit 32. The window area information is information in which rectangular information of the window area (position and size of the rectangle in the input image) is associated with the score.

  The area group storage unit 42 stores information on each area group generated by the area group generation unit 33. The area group information is information in which the label number (identification code) of the area group, the number of candidate areas belonging to the area group, and information on each candidate area are associated with each other. The candidate area information is rectangular information (rectangular position and size in the input image) and score. Of the stored area group information, the area group deletion unit 34 deletes the information that matches the deletion condition.

  The target area storage unit 43 stores information on a person area that is finally determined to have a person by the target area determination unit 35. The person area information is information in which rectangular information (rectangle position and size) of the person area in the input image is associated with the score, as in the candidate area information.

  The output unit 5 receives the result of the target area determination unit 35 and displays an alarm occurrence on the external display device such as a display together with the input image, or outputs an alarm signal such as sending an abnormality signal to the center device.

[Example of operation]
Next, the operation of the person detection device 1 will be described. FIG. 3 is a flowchart showing a schematic operation of the person detection apparatus 1. When an image is input from the image input unit 2 (step S10), the control unit 3 creates a reduced image by reducing the input image at a plurality of magnifications by the image reduction unit 30 (step S20). For example, as illustrated in FIG. 2, reduced images 110 and 120 are generated from the input image 100.

  The feature amount extraction unit 31 extracts a feature amount at each place in the image for each of the input image and the plurality of reduced images (step S30).

  The index value calculation unit 32 calculates a score corresponding to the window region set in each place in the image by using the feature amount extracted by the feature amount extraction unit 31 and the classifier stored in the index value calculation function storage unit 40. And stored in the index value storage unit 41 (step S40).

  The area group generation unit 33 extracts a person candidate area based on the score calculated by the index value calculation unit 32, generates an area group including a plurality of overlapping candidate areas, and stores the area group in the area group storage unit 42 ( Step S50).

  FIG. 4 is a schematic process flow diagram of the area group generation unit 33. The operation of the area group generation unit 33 will be described with reference to FIG.

Region group generation unit 33 refers to the index value storage section 41, extracts window regions score is first thresholds T ₁ or more as a candidate region (step S500).

  In FIG. 2, examples of candidate areas are indicated by solid-line rectangles having a size corresponding to the window area 101. In the image 100, candidate regions 102a and 102b are extracted around a small (far) person image on the left side. In the image 110, candidate areas 112a and 112b are detected around the central bus stop sign, and in the image 120, candidate areas 122a to 122d are extracted around a large (near) human image on the right side. As shown in FIG. 2, a plurality of overlapping candidate regions can be extracted for one image such as a person.

  FIG. 5 is a schematic image for explaining the subsequent process for the candidate area extracted by the area group generation unit 33. The image of FIG. 5 shows the same content as that shown in FIG. 2, and the image 130 of FIG. 5A displays the candidate areas of the images 100, 110, and 120 together on one image. Is. The image 130 is the same size as the input image 100, and the candidate areas 102a and 102b of the image 100 become candidate areas 131a and 131b on the image 130 at the same magnification. On the other hand, the candidate areas 112a, 112b, 122a to 122d in the reduced image are candidate areas 132a, 132b, 133a to 133d normalized to the magnification of the input image 100, respectively.

  The area group generation unit 33 rearranges the extracted candidate areas in descending order of the scores (step S501), and sets the label information for all candidate areas to a state indicating that the label number is not assigned (step S502).

  The area group generation unit 33 increments and sets the label number sequentially from 0. Therefore, the current label number is set to the initial value 0 (step S503).

  The area group generation unit 33 checks whether there is a candidate area to which the label number is not assigned (step S504). If there is an unassigned candidate area (in the case of “YES” in step S504), the unassigned candidate area having the highest score (candidate area A) is selected (step S505), The current label number is assigned (step S506).

  Then, a loop process (S507 to S511) is performed in which the candidate area A is used as a reference for comparison, and the candidate areas that have not been assigned labels are compared one by one. In the loop processing, candidate regions that are not selected as comparison partners are sequentially selected (step S507), the degree of overlap between the candidate region B selected as the comparison partner and the candidate region A is calculated (step S508), and the degree of overlap is calculated. Is larger than a predetermined group determination threshold value (step S509).

  The degree of overlap is calculated by, for example, (the area of the common area between candidate area A and candidate area B in the input image) / (the smaller of the areas of candidate area A and candidate area B in the input image). The Also, the degree of overlap is calculated by (area of the common area between candidate area A and candidate area B in the input image) / (area of the sum area of candidate area A and candidate area B in the input image). You can also. For example, the group determination threshold can be set to 0.5.

  If the degree of overlap is greater than the group determination threshold value (if “YES” in step S509), the same label number as the candidate area A is assigned to the candidate area B (step S510), and the process for the candidate area B is finished. The process returns to step S507. On the other hand, if the degree of overlap is equal to or less than the group determination threshold value (in the case of “NO” in step S509), the candidate area B is left unallocated with the label number, and the process for the candidate area B is finished and the process returns to step S507. Return.

  When the comparison with all the unallocated candidate areas for a certain candidate area A is completed, that is, when an unprocessed candidate area does not exist in step S507 and cannot be selected (in the case of “NO” in step S511), The current label number is incremented (step S512), the process returns to step S504, a new candidate area A is selected, and the processes of steps S505 to S512 are repeated.

  On the other hand, when all label numbers are assigned to the candidate areas, that is, when there is no unallocated candidate area (in the case of “NO” in step S504), the candidate areas to which the same label is assigned are defined as areas. The group is stored as a group in the area group storage unit 42 (step S513), the group generation process is terminated, and the process proceeds to step S60 in FIG.

  As described above, by setting a candidate area having a high score as the core of a group with priority, it can be expected that candidate areas relating to a plurality of adjacent persons are prevented from forming one group.

  As a result of the processing of the area group generation unit 33, for example, the candidate areas 133a to 133d in the image 130 in FIG. 5A become a group with a label number “0”, and the candidate areas 131a and 131b become a group with a label number “1”. The candidate areas 132a and 132b form a group with the label number “2”.

  Note that group generation based on the degree of overlap can also be performed by clustering using the centroid and dimensions of candidate regions as parameters.

  The area group deletion unit 34 deletes, from the area group storage unit 42, the area groups stored in the area group storage unit 42 that meet the above-described deletion conditions, that is, the requirements (R1) and (R2) (step S40). S60). FIG. 6 is a schematic process flow diagram of the area group deletion unit 34. The operation of the area group deletion unit 34 will be described with reference to FIG.

  The region group deletion unit 34 selects the region groups stored in the region group storage unit 42 as processing targets one by one and performs the processing of steps S601 to S603 (in the case of “YES” in step S600), unprocessed When there is no more area group, the group deletion process is terminated and the process proceeds to step S70 in FIG. 3 (in the case of “NO” in step S600).

For example, the region group deletion unit 34 first determines whether or not the region group to be processed satisfies the requirement (R1), that is, whether or not the number of candidate regions belonging to the group is equal to or less than a predetermined number threshold M. (Step S601). If the number is equal to or less than the number threshold M (in the case of “YES” in step S601), whether the score requirement (R2) is further satisfied, that is, the scores of all candidate regions belonging to the region group are the second threshold T _2. It is determined whether it is below (step S602). Note that as the second threshold value T ₂ has already been mentioned, it is a high-determined threshold than the first thresholds T ₁ for extracting a candidate region.

If the score is the second threshold value _{T 2} or less at step S602 (if at step S602 is "YES"), because the area group meets the requirements a deletion condition (R1) and (R2), the area group deletion The unit 34 deletes the area group from the area group storage unit 42 (step S603).

  On the other hand, if the requirement (R1) is not satisfied (in the case of “NO” in step S601) and the requirement (R2) is not satisfied (in the case of “NO” in step S602), the region group deletion unit 34 The area group is not deleted, and the process returns to step S600 to set the next area group as a processing target.

  The image 140 in FIG. 5B shows the processing result of the area group deletion unit 34 for the image 130. For example, assuming that the number threshold M is 2, the processing and effects will be described below.

The area group “0” composed of the candidate areas 133a to 133d does not satisfy the requirement (R1) because the number of candidate areas is 4, and thus is not deleted. The typical human region group sufficiently large candidate regions are extracted as may be appropriate to leave regardless of the relationship between the score and the second threshold value T ₂ of the candidate regions that constitute the group.

  On the other hand, the area group “1” composed of the candidate areas 131a and 131b attributed to the person satisfies the requirement (R1) because the number of candidate areas is 2, and the area group “2” composed of the candidate areas 132a and 132b attributed to the background. "" Also satisfies the requirement (R1) because the number of candidate regions is two. If the deletion condition is configured only by the requirement (R1) as in the prior art, the area group “2” caused by the background can be deleted, but the area group “1” caused by the person is also deleted.

However, by applying the present invention in which the requirement (R2) is included in the deletion condition, these can be appropriately distinguished. In other words, preferably not satisfy the requirement (R2) for the score of the candidate region 131a in a position to capture the true person area exceeds a second threshold value T ₂ are in the area group "1" due to the person, thus not deleted. On the other hand, the candidate region 132a in the area group "2", 132b meets the requirements score for due to the background becomes the second threshold value _{T 2} or less (R2), thus being removed.

  When the processing of the region group deletion unit 34 described above is completed, the target region determination unit 35 obtains a final person region and stores it in the target region storage unit 43 (step S70 in FIG. 3).

  An image 150 in FIG. 5C shows a processing example of the target area determination unit 35 for the image 140. The candidate areas 133a and 131a having the maximum score from the area groups “0” and “1” are person areas. As selected.

  If there is even one person in the image after the calculation of the person area in step S70 (in the case of “YES” in step S80), for example, the output unit 5 includes information on the detected person area and the person area. An abnormal signal including the detected input image is sent to the center device (step S90).

As described above, in the present invention described using the embodiment, the accuracy of the final detection result of the human region is improved by using the number of candidate regions included in the region group and the score of the candidate region together. Specifically, with the number of candidate regions included in the region group as a first index, one requirement (R1) of the region group deletion condition is that the number is below the threshold M. Further, as the higher second indicator threshold T ₂ which defines than the threshold T ₁ of the score when you find the candidate region, neither the score of the candidate region is deletion condition region group that below a second indicator Let it be another requirement (R2). By deleting a region group that satisfies both of these requirements (R1) and (R2), it is possible to accurately reduce background misdetection while preventing failure to detect the target and improve the accuracy of the final detection result. it can.

[Other Embodiments]
Hereinafter, only different points of the other embodiments of the present invention from the above embodiments will be described. Area group deletion unit 34 in the above embodiment has been deleted region group due to background based on the deletion condition with one second threshold value T ₂ but, as another embodiment, the second threshold value in the requirement (R2) T ₂ may be configured to change according to the number m of candidate regions included in the region group. For example, in the requirement (R2), the second threshold value T ₂ is set in a plurality of steps with respect to the number m of the following candidate area number threshold M.

Specifically, the setting values of a plurality of second threshold T ₂ set for the number m of the following candidate area number threshold value M is determined smaller the larger the number m. For example, for the number m, the range below the number threshold M is divided into the section b _H , the section b _M , and the section b _L in order from the top, and the set value of the second threshold T _{2 in} the section b is expressed as ρ (b). And T ₁ <ρ (b _H ) <ρ (b _M ) <ρ (b _L ). Incidentally, the number m included in each section may be one or plural.

Region group deletion unit 34 to the region group that meets the requirements (R1), determine whether they meet the requirements (R2) at the second threshold value T ₂ in accordance with the number m of candidate regions that belong to the area group To do. That is, the second threshold value T ₂ is determined to be lower as when the number m is large, the area group deletion unit 34 for the area groups, when the number m is not greater than the predetermined threshold value M, all attributable score of the candidate region is deleted as an area group due to the background what is the second threshold value T ₂ or less defined according to the number m.

For a high probability of a region group including the target region larger the number m of attributable candidate area, as the area group number m is larger attribution candidate region can be lowered a second threshold value T _2, whereby the detection of the target region The effect of preventing failure can be enhanced. On the other hand, since the number m of attribution candidate region is small region groups higher second threshold value T ₂ is applied, erroneous detection of the background area can be prevented to apply probability.

  In the above embodiment, the detection target is the whole body of the person, but a specific part such as a person's face or upper body may be the detection target, and various objects such as a vehicle or a sign may be the detection target. Various states such as poses and postures may be detected.

  DESCRIPTION OF SYMBOLS 1 Person detection apparatus, 2 Image input part, 3 Control part, 4 Storage part, 5 Output part, 30 Image reduction part, 31 Feature-value extraction part, 32 Index value calculation part, 33 Area group generation part, 34 Area group deletion part , 35 target area determination unit, 40 index value calculation function storage unit, 41 index value storage unit, 42 area group storage unit, 43 target area storage unit.

Claims (2)

A target detection device for detecting a target region where a predetermined target appears in an input image,
An index value calculation function for calculating an index value representing the likelihood that the target exists in a region of interest set in the input image using feature amounts extracted at various points in the input image is stored in advance. Storage unit
An index value calculation unit that sets the attention area at a plurality of positions in the input image and calculates the index value in the attention area by the index value calculation function;
A region composed of a plurality of candidate regions satisfying a predetermined overlapping relationship between the candidate regions and extracting the region of interest whose index value is equal to or greater than a predetermined first threshold as a candidate region An area group generation unit for generating a group;
In the region group, the number of candidate regions to which the group belongs is equal to or less than a predetermined number threshold value, and the index value of all the candidate regions to which it belongs is equal to or less than a second threshold value set higher than the first threshold value. An area group deletion section for deleting a certain thing,
A target region determination unit that determines the target region from each of the region groups that have not been deleted by the region group deletion unit;
An object detection apparatus comprising: When the number of the candidate regions belonging to the region group is equal to or less than the number threshold, the second threshold is set according to the number, and each setting value of the second threshold corresponds to the setting value The object detection apparatus according to claim 1, wherein the value is smaller as the number is larger.

Images