WO2012011068A1

WO2012011068A1 - Method and apparatus for identification and anthropometric classification of human body

Info

Publication number: WO2012011068A1
Application number: PCT/IB2011/053248
Authority: WO
Inventors: Elisabeth Duhamel
Original assignee: Cad Modelling Ergonomics S.R.L.
Priority date: 2010-07-22
Filing date: 2011-07-21
Publication date: 2012-01-26
Also published as: ITPI20100088A1; IT1401573B1

Abstract

A method for the identification and anthropometric classification of the human body comprises steps of acquiring images of the body by a scanning apparatus provided with a plurality of acquisition devices placed around the sample to be scanned, each acquisition device comprising a light source and a camera, processing the acquired images to form a point cloud and then a mathematical model of the human sample, identifying body potions, measuring peculiar sizes of the body portions and than assigning the body portions to anthropometric classes as a function of a comparison with data stored in specific databases. A system for identification and anthropometric classification comprises a three dimensional scanning apparatus particularly suitable to be self-operated and processing units which are able to process the acquired data and perform the assignment to anthropometric classes.

Description

Title of the invention

METHOD AND APPARATUS FOR IDENTIFICATION AND ANTHROPOMETRIC CLASSIFICATION OF HUMAN BODY

Field of the invention

The present invention relates to a method for performing automatic identification and anthropometric classification of a human body thanks to the detection of its volume, measuring peculiar size parameters and comparing them to reference values.

The present invention also relates to a tridimensional scanning apparatus having a plurality of acquisition devices suitable for autonomously and automatically scanning human bodies and automatically classifying the acquired bodies according to specific anthropometric classes.

Description of the Prior Art

The anthropometric classification of the human body is used for a number of purposes in different fields. In anthropology the anthropometric classification is useful for a statistical determination of models of development, modification and growth of a population. In the medical sector it can be useful for studies on obesity, both in childhood and in adultness. In product and process ergonomics the anthropometric classification can be widely used as it can be useful for designing all the tools, devices and machines which have to get in contact with humans (chairs, push-chairs, wheel-chairs, work spaces, toys, etc.). In the field of safety at work the anthropometric classification can be advantageously used for designing and producing personal protective equipments. Furthermore, the anthropometric classification can be useful to reconstruct crime scenes or crimes in general. Finally, in the field of clothing, are very important the studies on anthropometrics and are known various anthropometric classification theories which are applied for defining the size of clothes and to allow a correct sizing of clothes and dummies.

The development of equipments which are more and more reliable, cost saving and with a high resolution for 3D scanning of human body without contact allows looking for more and more effective methods for performing accurate and low cost anthropometric classifications.

A very efficient method for anthropometric classification is disclosed in EP 1 244 367 in which 3D scanning steps of a sample human body are followed by steps of processing the acquired data in which the values of linear and perimeter sizes are detected and compared with reference values of said sizes and the sample body is assigned to an anthropometric class according to specific algorithms.

The method disclosed in EP 1 244 367 allows a quite accurate classification as it makes use of ellipsoids intersecting the sample human body for detecting both linear and perimeter sizes which are then compared to corresponding numeric values stored in specific databases. As a result of the comparing step the human body is assigned to an anthropometric class. In addiction, the measured size values are added to the reference database and the definition of the anthropometric classes can be updated on a statistical base. Nevertheless, about this feature, the method disclosed in EP 1 244367 has one of its limits as the automatic updating of the definition of the anthropometric classes made without any external control can lead with time to definitions of the anthropometric classes which do not reflect the real anthropometric distribution of the population. Furthermore, in the classification process the different importance that the measured values (waist size, hip size, sleeve size, etc.) may have for determining to what anthropometric class the body belongs is not taken into consideration. Another limit of the method disclosed in EP 1 244 367, which is also present in all the method for anthropometric classification of the prior art, concerns the fact that only linear and perimeter size of the human body are measured and compared while shape features which are fundamental for classifying in morphologic families are not taken into consideration. Finally, one of the main limits of the above method concerns the type of 3D scanning apparatus used which has a certain structural intricacy, it is not easy to be autonomously used by the person to be scanned, it has a low detection accuracy due to the hardware configuration and to the use of a pivoting platform, it has high scanning times and it acquires a low amount of data.

In WO 2009/040368 it is disclosed an apparatus for tridimensional scanrung without contact in which a plurality of acquisition devices are mounted on independent tripods, each acquisition device being made of a light source and a camera which are able of rotating about a vertical plane. A preferable embodiment of the apparatus disclosed in WP 2009/040368 has two acquisition devices moiinted one on top of the other on the same tripod and there are three tripods placed at 120° around the object to be scanned. The apparatus allows a fast (about 4 seconds) and high accuracy acquisition and the object to be scanned do not need to be rotated or moved in any way. One of the main limits of the apparatus disclosed in WO 2009/040368, which is important mainly when it is used to scan people's bodies, is that it cannot be run directly by the person to be scanned. This can be very harmful for the commercial diffusion of this type of apparatus in the industry of clothing as many people are reluctant, for shame or other reasons, to be scanned if one ore more other persons need to be present for running the apparatus.

Summary of the Invention

It is an object of the present invention to propose a method for performing the identification and anthropometric classification of the human body with high accuracy and in a reliable way.

It is further object of the present invention to propose a method for identification and anthropometric classification which is comfortable to be used by the people working in the specific anthropometric classification sector and by the people to be scanned.

It is another object of the present invention to propose an apparatus for 3D scanning without contact of the human body which is particularly useful for performing identification and anthropometric classification in a comfortable and reliable way.

The above objects are attained by a method for the identification and anthropomorphic classification of human body based on acquiring images of the body surface of human samples and processing the acquired data through calculation algorithms suitable for creating three dimensional mathematical models of said human samples in which the acquisition is made without contact by a plurality of acquiring devices placed around said human sample, in which the method comprises steps of:

- acquiring images of a human sample by means of a plurality of acquisition devices in which each acquisition device is composed of a light source and a camera, each of said acquiring device rotating, during the acquisition, about a horizontal axis thereof;

- determining, from said images, a three dimensional point cloud corresponding to points of the human body surface of said human sample; - processing said point cloud for creating a three dimensional mathematical model of said human sample;

- identifying in said mathematical model portions of human body;

- extracting shapes, perimeters and linear dimensions of said portions of human body;

- comparing said shapes, perimeters and linear dimensions to reference shapes, perimeters and linear dimensions stored in databases associated to anthropometric classes;

- automatic assigning said portions of human body to respective anthropometric classes in which said automatic assigning is performed as a function of the results of said comparing and of a predetermined weight assigned to each of said shapes, perimeters and linear dimensions.

The method of the invention allows a fast and detailed acquisition of a number of body size parameters of human samples and it also allows an automatic, high accuracy anthropometric classification of body portions of said human samples.

Advantageously the step of assigning each of said portions of human body to an anthropometric class is followed by a step of manual re-assigning one or more portions of human body to anthropometric classes, said manual re-assigning producing an automatic updating of the parameters for perforrning said automatic assigning to anthropometric classes by modifying the importance given to each of said shapes, perimeters and linear dimensions.

According to the method of the invention it is possible to manually correct the automatic classification, most of all when a body portion is close to match more than one anthropometric class. Upon manual correction of a classification according to the method of the invention the calculation algorithm is updated just like a self-learning procedure which continuously increases the accuracy of the classification.

Advantageously the step of acquiring images is performed in a time interval lasting about 4 seconds.

Still advantageously the three dimensional point cloud calculated on the basis of the acquired images has a bigger density in a zone corresponding to a front and central section of said human sample. The above objects are also attained by a system for the identification and anthropometric classification of human body comprising:

- an apparatus for three dimensional scanning without contact comprising:

- a plurality of acquiring devices each composed of a light source and a camera;

- at least three independent holding tripods, each supporting at least one of said acquiring devices;

- actuating means for rotating said light sources and said cameras of each acquiring devices about a horizontal axis thereof;

- at least a control unit for controlling said light sources, said cameras and said actuating means;

- at least a processing unit for converting image4s acquired by said cameras into a three dimensional mathematical model of said human sample, for identifying in said three dimensional mathematical model portions of human body, and for calculating numerals corresponding to measurements of said human sample,

in which said system is characterized in that it comprises:

- memory means comprising at least a database of shapes, perimeters and linear dimensions;

- at least a further processing unit for processing said numerals comprising algorithms for:

* comparing said shapes, perimeters and linear dimensions to reference shapes, perimeters and linear dimensions stored in said memory means and associated to anthropometric classes;

" assigning said portions of human body to respective anthropometric classes;

and in which said apparatus for three dimensional scanning comprises means for remote and/or delayed activation.

The presence of two separate processing units, each deputed to specific works, and the presence of means for remote and/or delayed activation render the system very efficient and comfortable to be used. Brief Description of Drawings

These and other characteristics of the invention will become more easily comprehensible from the following description of a preferred embodiment of the invention, given as not limiting example, with reference to the enclosed drawings, in which:

• Figure 1 shows a flow chart of the steps of a method according to the invention;

• Figure 2 shows a perspective view of an apparatus according to the invention;

• Figure 3 shows a side view of an acquisition device comprised in an apparatus according to the invention.

Description of the Preferred Embodiments

With reference to Fig. 1, it is shown a flow chart useful as an example of the main steps of a method according to the invention.

After a step of fuming on and calibration of the system, a human sample enters the acquisition place and, once he assumed the correct position he autonomously starts the acquisition, step 110. The acquisition devices, rotating in a vertical plane acquire images of the body of the human sample, step 120. The acquisition step 120, in which the human sample has to keep unmoving in the acquisition position, lasts no more than 4 seconds, and then the human sample can leave the acquisition position and get dressed again. The acquired images are transmitted to a first processing unit which is able to define a three dimensional point cloud, step 130, corresponding to point of the surface of the body of the scanned human sample. The above point cloud can be displayed by visualization means of the processing unit for an early verification that the images have been correctly acquired. Then, through a common three dimensional interpolation algorithm from the point cloud is obtained a three dimensional mathematical model, step 140, of the body of the human sample.

At this point, a specific algorithm is able to detect body portion, step 150, in the mathematical model and split the mathematical model in various portion corresponding to parts of the human body (for instance, legs and feet; arms and hands; hips, waist and stomach; chest and head). A further specific algorithm extracts, for each portion of human body, or even in an aggregate mode for several portions, peculiar shapes, perimeters and linear dimensions, step 160. The above shapes, perimeters and linear dimensions can be differently defined as a function of the field of application in which the anthropometric classification is performed. In general, for each mathematical model representing a detected body portion, or even for a composition (aggregation) of body portions, are measured linear sizes or perimeters, but also shape parameters that, for a same perimeter value, strongly distinguish the human sample from an anthropomorphic point of view. In order to have clearer the importance of such shape parameters it is sufficient to think to the great anthropomorphic differences that could have people with a same waist size or chest size.

Then, the shapes, perimeters and linear dimensions measured are compared to corresponding reference shapes, perimeters and linear dimensions, step 170, stored in specific databases, step 180, of anthropometric classes made on a statistical base. The above comparison is made by using an algorithm of classification which is realized according to the theories which are the basis of the definition of anthropometric classes.

Basically, for each morphologic family (class) there is in the database a specific set of reference sizes, both linear and shape dimensions, each of them having a determined weight which is taken into account by the classification algorithm, and each reference size is an average value of the real values of that size assigned to the specific morphologic family.

Upon comparison, to each shape, perimeter or linear dimension is assigned a matching percentage with the specific anthropometric class. In addiction, as above stated, each shape, perimeter or linear dimension has its weight, that is its importance, for determining the assignment of the body portion to a specific anthropometric class. Practically, for determining the anthropometric class to which a body portion belongs are considered both the difference between the value of the measured dimension and the reference value of the dimension (here comprised the similarity between the reference and the measured shapes), and the weight (importance) of the shapes, perimeters and linear dimensions.

The algorithms, whose way of working has been briefly explained above, defines the assignment of each body portion to a specific anthropometric class, step 190, and it also determines a percentage of match with the assigned class. Obviously, that body portion will have a lower matching percentage with the other anthropometric classes.

In some cases, the difference between the percentage of match with the assigned anthropometric class and the percentage of match with other classes could be very little. In these cases, according to the method of the invention, the user interface gives the user the chance of making a manual re-assignment of the body portion to a different anthropometric class, step 200, since a user, thanks to its skill and for various reasons (for instance a visual comparison between the main shapes of the human sample and the reference shapes of the various anthropometric classes), could find more appropriate the assignment to a different class. If no re-assignment takes place the class is definitively the previously assigned class. The values of the various measured sizes of the human sample are used to update the database, step 210. Differently, when a manual re-assignment is made for one or more body portions the new classes manually chosen by the user are the classes of assignment, 220, and the system automatically updates the classification algorithm, step 230. In fact, the system takes into consideration the manual re-assignment and consequently modifies according to specific parameters the weight of some shapes, perimeters and linear dimensions. Basically in the step 230 of updating the classification algorithm the system performs a kind of self-learning procedure which allows a continuous improvement of the classification algorithm.

Obviously, the classification can be made in detail by assigning each body portion to an anthropometric class, or it can be made in an aggregate mode by identifying a single anthropometric class for the entire human sample or a class for the upper portion and a class for the lower portion of the body.

The above disclosed method is advantageously performed through an apparatus comprising devices for three dimensional acquisition without contact and processing units specifically designed and intended to the identification and anthropometric classification of human samples. The main features and working principles of such an apparatus are disclosed in WO 2009/040368 and they are briefly mentioned in the following with particular reference to an embodiment which suits very well the method of the invention. With reference to figs. 2 and 3 it is shown an apparatus for three dimensional scanning without contact and the anthropometric classification of human samples comprising three support tripods 10 placed at about 120° each other along a circle, C, whose center represents the acquisition position where stands the human sample, U. The tripods 10 are completely independent each other so that they can be placed one at a time without any risk of misalignment which could compromise the procedure of acquiring the human sample U. In fact, the accuracy of the acquisition is assured by a preliminary calibration which is very easy and rapid. In addiction, the angle between the tripods could be different from 120°. In fact, advantageously the angle between the tripods facing the human sample U is reduced and the angle between those tripods and the back tripod is increased. Thus implies the acquisition of a greater amount of data for the front surface of the body with respect to the back one.

On each tripod 10 are mounted two acquisition devices, 20, vertically aligned along the tripod itself. Each acquisition device 20 is rotatably bound to the tripod 10 for rotating about a horizontal axis, 31, substantially tangent to the circle C, thanks to drive means, 30. Control means, 40, controls the acquisition devices 20 and the drive means 30 thereof.

As it can be seen in fig. 2 each acquisition device 20 is composed of a light source, 21, and a camera, 22, vertically aligned along a support arm, 23 rotating about the axis 31. The light source 21 and its camera 22 are disposed with their axis incident according to an angle a that in this embodiment is a constant angle during the acquisition procedure. When the acquisition procedure starts the light source 21 lights with a beam of light the human sample U and the camera 22, preferably a digital camera, acquires images in sequence of the light beam on the human sample. During the acquisition, each acquisition device rotates about its axis 31. Thanks to the presence of two acquisition devices 20 on each tripod 10, to the presence of three tripods, and to the kind of digital cameras 22 employed, the apparatus acquired in about 4 seconds an amount of data which are sufficient to realize a mathematical model of the surface of the body of the human sample in a very detailed and accurate way.

In addiction, the apparatus of the invention is designed and sized so that it is able to perform an accurate acquisition of any type of human sample. In fact, the height of the various light sources 21 and the cameras 22 thereof, the height of the rotation axis 31, and the angles of start and end rotation of the acquisition devices are designed for eliminating dead acquisition areas. In particular, it has been seen that, mainly with human sample having a big stomach, it is preferable that the lower acquisition devices makes the acquisition from a lower point of view. At this aim, with respect to the embodiment shown in figs. 2 and 3, it is sufficient to overturn the lower support arms 23 so that the light sources 21 and the cameras 22 they support are moved downwards and the start and end acquisition angles are accordingly modified upwards. In this way it is possible to have the proper light directed towards critical areas such as the downward directed portion of a prominent stomach.

The mathematical model of the surface of the body of the human sample U is calculated by a processing unit, 50, in which the data acquired by the cameras 22 are processed and put together. The incidence angle of the light source 21 and the camera 22 thereof is used in a specific algorithm of the processing unit 50 together with trigonometric formulas for generating points in a three dimensional reference which corresponds to points of the surface of the body of the human sample. The data coming from all the acquisition devices are pt together in order to form a single point cloud in a three dimensional reference corresponding to points of the surface of the body of the human sample U. As described in the disclosure of the method of the invention, further algorithms in the processing unit 50 are able to generate a three dimensional mathematical model, to identify body portions and to measure specific shapes, perimeters and linear dimensions of such body portions.

With respect to what disclosed in WO 2009/040368 the acquisition apparatus of the invention has some innovative peculiar features suitable to overcome limits which are relevant in the specific field of use of the present invention, that is of the identification and anthropometric classification of human samples.

First of all, the apparatus of the invention is provided with means for remote and/or delayed activation, 60, thanks to which the human sample U is able to autonomously run the apparatus. It is very important in the specific field of the invention the presence of means for the remote activation, irrespective whether they are of a wireless as shown in fig. 2 or cable type, or means for a delayed activation such a timer able to give the human sample U the chance of taking the right position in the acquisition area after having given a command of start scanning. In fact, without the above means the commercial diffusion of the apparatus could be limited due to the fact that many people are reluctant to stand almost nude (for a correct scanning the human sample has to wear at most only pants or a very tight and adherent suit) in front of a runner of the apparatus.

Furthermore, differently from all other kind of apparatus used in the prior art for similar purposes, in the apparatus of the invention the acquisition gain, the light and the contrast of the acquired images can be adjusted in a very simple and rapid way, by gain, light and/or contrast set means, in order to overcome troubles rising in specific conditions (dark skin, a hairy body) in particular when a laser light source is used.

In addiction, the apparatus of the invention is provided with a further processing unit, 70, physically distinct from the processing unit 50 previously mentioned.

The processing unit 70 comprises or is operatively connected to memory means comprising the database where are stored the data about sizes and shapes which are relevant for the anthropometric classification and it also comprises the algorithms of the comparing and classification steps as well as the algorithm for updating the previous ones.

The presence of two different processing units, 50 and 70, is particularly useful since they perform different works. The processing unit 50, which is operatively connected to the acquisition devices has to be connected in the fast and close possible way to these ones and it needs good computational capacities. The second processing unit 70 could have much lower computation capacities since it just need to be able to access and manage databases.

The characteristics and advantages of the method and apparatus above described remain unchanged also in presence of modifications or with different embodiments.

As concerns the method for identification and anthropometric classification of the present invention it will be clear to people skilled in the sector the acquisition of images of the human sample could be performed according to procedures which are different from what above disclosed. The number of body portions identified can be defined according to the specific needs, as well as the kind of dimensions and shapes measured could be different.

These and other versions or modifications could be applied to the method and apparatus according to the invention, still remaining within the protective scope defined by the following claims.

Claims

1. Method for the identification and anthropomorphic classification of human body based on acquiring images of the body surface of human samples and processing the acquired data through calculation algorithms suitable for creating three dimensional mathematical models of said human samples in which the acquisition is made without contact by a plurality of acquiring devices placed around said human sample, characterized in that the method comprises steps of:

acquiring (120) images of a human sample (U) by means of a plurality of acquisition devices (20) in which each acquisition device is composed of a light source (21) and of a camera (22), each of said acquiring devices rotating, during the acquisition, about a horizontal axis (31) thereof;

determining (130), from said images, a three dimensional point cloud corresponding to points of the body surface of said human sample (U);

processing (140) said point cloud for creating a three dimensional mathematical model of said human sample (U);

identifying (150) in said mathematical model portions of human body;

extracting (160) shapes, perimeters and linear dimensions of said portions of human body;

comparing (170) said shapes, perimeters and linear dimensions to reference shapes, perimeters and linear dimensions stores in databases (180) associated to anthropomorphic classes; automatic assigning (190) said portions of human body to respective anthropomorphic classes in which said automatic assigning is performed as a function of the results of said comparing and of a predetermined weight assigned to each of said shapes, perimeters and linear dimensions.

2. Method according to claim 1 characterized in that said step of assigning (190) each of said portions of human body to an anthropomorphic class is followed by a step of manual reassigning (200) one or more portions of human body to anthropomorphic classes, said manual re-assigning (200) producing an automatic updating (230) of the parameters for performing said automatic assigning (190) to anthropomorphic classes.

3. Method according to the previous claim characterized in that said step of automatic updating (230) changes said predetermined weight assigned to each of said shapes, perimeters and linear dimensions.

4. Method according to any preceding claim characterized in that said step of acquiring images (120) is performed in a time interval lasting about 4 seconds.

5. Method according to any preceding claim characterized in that said three dimensional point cloud has a bigger density in a zone corresponding to a front and central section (bust) of said human sample (U).

6. System for the identification and anthropomorphic classification of human body comprising:

apparatus for three dimensional scanning without contact comprising:

a plurality of acquiring devices (20) each composed of a light source (21) and a camera (22); - at least three independent holding tripod (10), each supporting at least one of said acquiring devices (20);

- actuating means (30) for rotating said light sources (21) and said cameras (22) of each acquiring device (20) about a horizontal axis (31) thereof;

- at least a control unit (40) for controlling said light sources (21), said cameras (22) and said actuating means (30);

- at least a processing unit (50) for converting images acquired by said cameras (22) into a three dimensional mathematical model of said human sample, for identifying in said three dimensional mathematical model portions of human body, and for calculating numerals corresponding to measurements of said human sample,

characterized in that it comprises:

memory means comprising at least a database (180) of shapes, perimeters and linear dimensions;

at least a further processing unit (70) for processing said numerals comprising algorithms for:

- comparing said shapes, perimeters and linear dimensions to reference shapes, perimeters and linear dimensions stores in said memory means and associated to anthropomorphic classes;

- assigning said portions of human body to respective anthropomorphic classes,

wherein said apparatus for three dimensional scanning comprises means for remote and/or delayed activation (60).

7. System for the identification and anthropomorphic classification of human body according to the previous claim characterized in that it comprises adjusting means for adjusting gain, lightness and contrast of said cameras (22).