Method and system for detecting a construction waste sorting error

Abstract

Computer-implemented method and system for detecting a construction waste sorting error on a construction site including at least one dumpster for receiving construction waste, the method comprising the step of obtaining image data representing a content of a first dumpster for receiving construction waste of a first type.

Description

[0001] METHOD AND SYSTEM FOR DETECTING A CONSTRUCTION WASTE SORTING ERROR

[0002] Field of the Invention

[0003]

[0001] The present invention generally relates to a computer-implemented method and system for detecting a construction waste sorting error on a construction site including at least one dumpster for receiving construction waste.

[0004] Background of the Invention

[0005]

[0002] Waste recuperation on building sites is known to be an important issue. Very different building materials may be used and disposed of, often by working people from different companies. Construction waste may for example include wood, metal, plaster, building rubble, various types of insulating materials, and many other types of material. When all waste is collected in a single container, it may become relatively difficult and costly to sort it out by type, for example for recycling purposes. Even when separate containers per material are present on a building site, workers may be tempted to throw away construction waste in the nearest container rather than in the correct container. As a consequence, the content may still need further sorting, which may be technically relatively difficult and costly.

[0006]

[0003] Systems including Al-based waste segregation are known in the context of household waste, in particular in closed dumpster, and is generally limited to the recognition of homogeneous and small sized waste, such as paper, plastic and / or metal. However, such systems cannot be applied to building sites for several reasons. Camera surveillance on building sites has proven to be relatively cumbersome to provide. Very often, only limited power and / or no or hardly any internet connection is available on site. Dumpsters on a building site are generally open, which can complicate image recognition given the potentially variable environmental conditions, such as light, dust and / or water. Next, a surveillance system should be relatively easy to install and uninstall given the short-term use of the system on a construction site. Moreover, an implementation of such a surveillance system should be very flexible since every construction site is different and may have to deal with different types of material flows. When a system and / or method includes image recognition based on machine learning or Al, for example based on semantic segmentation which is a relatively complex Al model, it may take a relatively long time to train the system to recognize all the different materials which may potentially be present on a building site. At the same time, such a system may not be very robust. Even if a system can be trained to a relatively high level of reliability for a specific site only, said trained system may then not be suitable for another building site where different types of materials may be present.

[0007] Summary of the Invention

[0008]

[0004] It is therefore an aim of the present invention to solve or at least alleviate one or more of the above-mentioned problems. In particular, the invention aims at providing a relatively simple and robust computer-implemented method for detecting a construction waste sorting error which is versatile and can be used on different construction sites.

[0009]

[0005] To this aim, according to a first aspect of the invention, there is provided a computer-implemented method for detecting a construction waste sorting error on a construction site having the features of claim 1. In particular, the method comprises the following steps. Image data representing a content of a first dumpster for receiving construction waste of a first type are obtained. The image data may be obtained directly from a camera or may be obtained from any image data storage medium. The image data represent the content of a first dumpster, which is positioned on a construction site, and which is configured to receive construction waste of a predetermined first type, for example to receive wood or any other type of construction waste which is present on said construction site. Then a first binary classifier that is individually trained exclusively on detecting a first type of construction waste checks a presence of said first type of construction waste in the content of the first dumpster, as represented by the image data, thereby obtaining a first binary presence result. As an example, the first binary classifier is an individually trained convolutional neural network that is trained only or exclusively on detecting a first type of construction waste and only checks whether or not wood is present in the image data, for example by any known Al-based image classification method. Different Al-based classification methods may be considered such as for instance convolutional neural networks, transfer learning with pre-trained models, support vector machines (SVM) with feature extraction, or alike. The result of said presence check is either ‘yes’ meaning that wood is present in the image data, or ‘no’ meaning that no wood at all is present in the image data. In the first case, when there is a positive detection of wood, it does not imply that only wood is present in the image data. The image data may represent wood and any other additional type of construction waste. Therefore, the method further comprises the step of checking by at least a second binary classifier a presence of at least a second type of construction waste in the content of the first dumpster, thereby obtaining an at least second binary presence result. This second binary classifier is also individually trained, meaning that the Al-based image classification method is exclusively trained to detect one type of construction waste. As an example, the second individually trained binary classifier only checks whether or not metal is present in the image data. The result of said presence check is again a binary result, either ‘yes’, metal is present in the image data, or ‘no’, no metal has been detected in the image data. Again, the presence of metal does not exclude the presence of any other type of construction waste. The first binary classifier and the at least second binary classifier may preferably be configured to perform the presence check in parallel on the same image data. Alternatively, the presence checks by the first binary classifier and by at least a second binary classifier may be performed sequentially. Finally, a detection of a construction waste sorting error is output when said first binary presence result and said at least second binary presence result include at least two positive presence results. As an example, if the presence of wood is detected by the first binary classifier in the image data representing the content of the first dumpster and metal is detected by the at least second binary classifier in said same image data, it means that wood and metal have been mixed in the same dumpster, which triggers an output as a construction waste sorting error. The output may include outputting a control signal to a warning system, for example a visual warning and / or acoustic warning system, showing a warning on a display, sending warning messages to mobile devices within a predetermined range or any combination thereof, or any other warning output known to the person skilled in the art. The output may be provided by any known and suitable data selector or combinational logic circuit, such as a multiplexer using a XOR operation.

[0006] As a result, the present method can provide a versatile solution for construction waste management on a construction site. Since the Al-based algorithm is relatively lightweight, it can be trained relatively quickly onsite based on types of construction waste which are present on and specific to the site. When an Al-based classification algorithm is individually or exclusively trained on a single type of object, or in our case a single type of construction waste, the Al-based algorithm or model can be trained relatively fast, and the classification tasks can be performed using little computational resources. For instance, a convolutional neural network CNN that is trained to detect a single object in an image, requires much less training time compared to a CNN that is trained to detect the presence of multiple object types. Moreover, execution of the algorithm can be performed on relatively simple hardware without requiring an internet connection onsite and with only very little power needed which can be provided by for example small batteries. Since also the training of the Al-based classification model requires less computing resources, the training or retraining of the model can be performed on-site. This aspect provides the additional benefit that the model is trained on images that match the actual illumination and environmental conditions where the model will be used.

[0010]

[0007] The first binary classifier, as well as the at least second binary classifier, can be trained by providing manually labelled images to the classifier, preferably including images of three types, even if the classifier is trained to provide only a binary presence result: images with only one type of waste, for example wood only, images with the first type of waste and any other type of waste, for example wood mixed with metal, and images without the first type of waste. In a next step, similar labelled images may be provided in various potential environmental conditions, for example a content of the dumpster under variable luminosity, in the rain, covered by dust or snow or any other potential (weather) condition. Since the training is only a binary training, the training can be performed relatively quickly.

[0011]

[0008] The method can preferably further comprise the step of receiving binary feedback when a detection of a construction waste sorting error is erroneously output and the step of retraining the first binary classifier and the at least second binary classifier based on said binary feedback. In this way, the output of the method can be improved continuously. As an example, a detection of a waste sorting error may have been output for a combined detection of wood and metal. When a user checks the content of first dumpster and / or the image data representing the content of the first dumpster, the user may conclude that no metal is present. The user may provide this feedback, which is generally known as a false positive, to the system, in the form of binary feedback, for example whether the output detection of a waste sorting error is correct or not, for example through any known interface, such as a screen, a mobile device, or any dedicated feedback interface, such as a button to approve or disapprove of the construction waste sorting error. The method then preferably uses this feedback received from said interface as additional training data for the method such that the first and at least second binary classifiers can be retrained while the method and / or system is in use, for example on a construction site. This type of feedback can be used relatively easily for retraining since it is clear from the erroneous output which classifier has made an error.

[0012]

[0009] The method can preferably further comprise the step of receiving binary feedback when a non-detection of a construction waste sorting error is erroneously output and retraining the first binary classifier and the at least second binary classifier based on said binary feedback. When no waste sorting error is detected, the method may optionally include a positive output, in other words, outputting an absence of detection of a construction waste sorting error when the first binary presence result and said at least second binary presence result include only one positive presence result. Such a positive output may for example include a green light, a thumbs up sign on a display, or any other indication that no waste sorting error has been detected. If such a positive output turns out to be wrong, in particular when a user does detect a waste sorting error without the error being output or with a positive output, which is generally known as a false negative, then this feedback may also be provided to the method, for example through a display, as binary feedback, telling the method that a construction waste sorting error has been detected in spite of an absence of output of said detection or in spite of positive output. However, in this case, to retrain the first binary classifier and the at least second binary classifier more efficiently, additional info may be needed to identify which classifier has made an error since the error can come from any of the binary classifiers. Additional information, for example on the type of waste which was manually detected in the content of the first dumpster and / or in the image data representing said content of the first dumpster, may for example be provided via any suitable interface, or for example via a dedicated app on a mobile device.

[0013]

[0010] The method can advantageously comprise the step of training said first binary classifier and the at least second binary classifier using labelled image data representing construction waste of said construction site. In this way, the method can be adjusted to and focussed on the specific waste which is present on that construction site. This is made possible because the training is relatively fast thanks to the relatively simple algorithm, based on a combined use of a plurality of binary classifiers. The training of the plurality of classifiers may even be performed onsite.

[0014]

[0011] The method may further include the step of checking by a further classifier a presence of a further type of construction waste which is different from the first type and the at least second type, thereby obtaining a further binary presence result. The further classifier can preferably be a generally available classifier, not necessarily a binary classifier, which is for example trained to recognize a limited range of additional waste which is often present on a construction site, but which is not construction waste in itself. For example, the further classifier may be trained to detect a banana peel, or other food rests, or bottles in glass or plastic, or any known type of waste. Even if the classifier may be a general classifier, trained to detect for example a limited number of general waste items, the result is preferably binary such that it can be treated in combination with the first and at least second binary presence results, in particular by a XOR operator. The output may still be such that a waste sorting error is output as detected when there are at least two positive binary presence results, for example wood and a banana peel.

[0015]

[0012] It is preferred that the method further comprises the step of storing the obtained image data when detecting a construction waste sorting error, i.e. when a mix of different construction waste types is detected and output. Storing said image data can help a user in detecting the wrongly sorted construction waste and / or in identifying an origin of the wrongly sorted construction waste, which may be important on a construction site with respect to responsibilities and cost allocation, in particular when a user ignores any warning signal. It is preferred that image data are only stored when an error is detected to avoid the need for a large data storage. The data may be stored locally or remotely, for example in a cloud system if any data connection is available onsite.

[0016]

[0013] In an advantageous way, the obtained image data may further represent a content of at least a second dumpster for receiving construction waste of at least a second type. The method may then further comprise a step of selecting a single dumpster among the first dumpster and the at least second dumpster represented in the image data before performing the remaining steps of the method on each dumpster separately. The method may thus be repeated per dumpster, even if the image data can show content of a plurality of dumpsters. In this way, only one camera is needed, which may then be configured to provide a single image representing the content of a plurality of dumpsters. Alternatively, one camera may be provided per dumpster such that image data only represents the content of a single dumpster.

[0017]

[0014] Said first type and said at least second type of construction waste may preferably include one of wood, metal, plaster, or building rubble. More preferably, the method may for example be performed for four types of construction waste by four binary classifiers, each being trained to check the presence of one of said four types of construction waste. A presence of other types of construction waste may be checked as well, for example insulation materials, glass wool, polyethylene (PE) foam, styrofoam, or any other type of construction waste, each by a dedicated binary classifier. More or fewer binary classifiers may be used to an advantage as well.

[0018]

[0015] The obtained image data may preferably be a greyscale image. More preferably, the received image is a black and white image. If the image is a colour image, the image may be transformed into a greyscale image. In this way, an influence of colour- related issues in the recognition of various types of waste may be decreased, resulting in improved reliability of the results of the binary classifiers.

[0019]

[0016] The method may advantageously further comprise the step of obtaining further image data representing a content of said first dumpster for receiving construction waste of said first type and the step of applying a mask to at least part of said further image data when no construction waste sorting error has been output for a first set of image data representing a content of said first dumpster. In other words, when said first binary presence result and said at least second binary presence result include only one positive presence result, it means that no construction waste sorting error has been detected. When further image data representing a content of said first dumpster for receiving construction waste of said first type are obtained again, in particular, after further construction waste has been added to the first dumpster, a mask may be applied to these further image data to mask the already checked construction waste. The method may then be applied to subtracted image data resulting from a subtraction between the first set of image data and the further image data of the same first dumpster while the binary presence results obtained on the first set of image data and on the subtracted image data may be combined. Such a mask may at least partially obviate potential problems in image recognition, in particular when dust has covered the construction waste in between the capturing of the first set of image data and the further image data.

[0020]

[0017] According to a further aspect of the invention, there is provided a controller, a computer program product and a computer readable storage medium having the features of claims 10 - 12 respectively. They can provide one or more of the above- mentioned advantages.

[0021]

[0018] According to a further aspect of the invention, there is provided a system for detecting a construction waste sorting error on a construction site having the features of claims 13 - 15. Such a system can provide one or more of the above-mentioned advantages. As explained before, the system can include a single camera configured to capture image data of a content of one or more dumpsters, or alternatively, the system can include a plurality of cameras, each being configured to capture image data of a content of one or more dumpsters. The system can preferably include a combined warning system, such as a flashing light combined with an acoustic alarm. Many other warning devices may be used in the system, as will be clear to the person skilled in the art. The system may preferably include interfacing means which may include binary feedback means, such as for example two buttons, for example a green button for ‘correct’ and a red button for ‘wrong’. The interfacing means may further include means allowing non-binary feedback, such as via a display, or via a mobile application. The warning devices, as well as the interfacing means, are preferably compatible with an outside working environment of a construction site. Brief Description of the Drawings

[0022]

[0019] Fig. 1 shows a schematic view of a preferred embodiment of a system for detecting a construction waste sorting error on a construction site according to an aspect of the invention;

[0023]

[0020] Fig. 2 shows a schematic view on a preferred embodiment of a method for detecting a construction waste sorting error on a construction site according to a further aspect of the invention;

[0024]

[0021] Fig. 3a and 3b show schematic views of a further preferred embodiment of the method of Figure 2;

[0025]

[0022] Fig. 4a and 4b show schematic views of another preferred embodiment of the method of Figure 2; and

[0026]

[0023] Fig. 5 shows a suitable computing system comprising circuitry enabling the performance of steps of embodiments of the method of Figure 2.

[0027] Detailed Description of Embodiment(s)

[0028]

[0024] Figure 1 show a schematic view of a preferred embodiment of a system 1 for detecting a construction waste sorting error on a construction site according to an aspect of the invention. The construction site includes at least one dumpster 2, preferably a plurality of dumpsters 2, each of them being configured to receive construction waste of a respective predetermined type 3. As an example, a first dumpster 2a may be configured to receive plaster 3a, represented as triangles, a second dumpster 2b may be configured to receive building rubble 3b, represented as bricks, a third dumpster 2c may be configured to receive construction wood 3c, represented as cylinders, and a fourth dumpster 2d may be configured to receive metal 3d. Said dumpsters 2 can preferably be recipients having an open top side, for example containers or big trash bags. Said dumpsters 2 may each include a sign indicating which type of construction waste is to be dumped into said respective dumpster 2. The system 1 for detecting a construction waste sorting error comprises at least one camera 4 to capture image data of a content of at least one dumpster 2. The system 1 further comprises a controller 5 configured to perform the computer-implemented method as claimed and which will be described with respect to Figure 2. The controller 5 may for example be integrated into the camera 4. In other words, the method may be run on a relatively small graphical process unit (GPU), for example on a 4GB small GPU, such as a NVIDIA Maxwell architecture or Jetson Nano with 128 and a quadcore of 1.47 Ghz. Such a system is a low-energy system and can be relatively autonomous. The system 1 can further include a warning device 6, such a flashing light or a siren configured to visually and / or acoustically output a warning signal when triggered by the controller 5. The system 1 further comprises interfacing means 7 configured to allow a user to provide feedback, in particular on the correctness of the warnings emitted by the system 1. Said interfacing means 7 may comprise binary feedback, for example with buttons indicating whether or not a user has manually detected a sorting error, as will be explained with respect to Figures 4a and 4b. The interfacing means 7 may additionally or alternatively include a mobile application configured to receive warnings and / or provide feedback means. The interfacing means 7 and / or the camera 4 may optionally be cloud-connected 8, for example to store image data when a construction waste sorting error has been detected.

[0029]

[0025] Figure 2 shows a schematic view on a preferred embodiment of a method 10 for detecting a construction waste sorting error on a construction site according to a further aspect of the invention. In a first step 11 , image data are obtained, the image data representing a content of a first dumpster 2a for receiving construction waste of a first type 3a. Said image data may, but need not, be obtained directly from a camera 4. When the obtained image data represent a content of at least a second dumpster 2b, or a content of a plurality of dumpsters, a content of for example four dumpsters 2a, 2b, 2c, 2d, each being configured to receive construction waste of a respective predetermined type 3a, 3b, 3c, 3d, as shown in Figure 1 , then the method preferably further comprises a step of selecting a single dumpster among the first dumpster 2a and the at least second dumpster 2b, 2c, 2d represented in the image data. In that case, the method can be performed on each dumpster separately, sequentially or in parallel. The image data can preferably be greyscale image data, or even black and white image data, either originally or transformed into greyscale or black and white.

[0030]

[0026] In a next step 12, the obtained image data are analysed by a first binary classifier 13a to check a presence of said first type of construction waste 3a in the content of the first dumpster 2a as represented in the image data, based on known image recognition techniques, thereby obtaining a first binary presence result. In other words, the first binary classifier 13a has been trained to check whether or not a predetermined type of construction waste is present. The result can be a positive result, when the predetermined type of construction waste is present, in combination with other types of waste or not, or the result can be a negative result, when no waste of said predetermined type has been detected. The obtained image data are analysed as well by at least a second binary classifier 13b to check a presence of at least a second type of construction waste in the content of the first dumpster 2a, thereby obtaining an at least second binary presence result. More preferably, the image data may be analysed by a plurality of binary classifiers, for example four binary classifiers 13a, 13b, 13c, 13d, each binary classifier being trained to check the presence of a predetermined type of construction waste. In this way, a plurality of binary presence results is obtained for said image data representing the content of the first dumpster 2a. Said binary presence results are then transferred to a data selector 14, for example a multiplexer, configured to provide an output based on said plurality of binary presence results.

[0031]

[0027] In a final step 15, a detection of a construction waste sorting error is output when said first binary presence result and said at least second binary presence result include at least two positive presence results. When there is only one positive presence result, then it is assumed that there is no construction waste sorting error. This final step 15 may for example be performed by a XOR operation on a multiplexer. Said output may for example trigger a warning device 6 to output a visual and / or acoustic warning signal when a construction waste sorting error has been detected, as will be explained in more detail with reference to Figures 3a and 3b.

[0032]

[0028] Figures 3a and 3b show schematic views of a further preferred embodiment of the method of Figure 2. In particular, an outcome of step 12 as well as step 15 have been illustrated. In Figure 3a and 3b, dumpster 2a is configured to receive construction waste of a first type 3a, represented as triangles. Image data obtained by the camera 2 representing a content of said dumpster 2a have been analysed by four different binary classifiers 13a, 13b, 13c, 13d, each binary classifier being trained to check the presence of a predetermined type of construction waste, as for example represented by an arch, a cylinder, a bar and a triangle. These different types of construction waste may, but need not, correspond to the types of waste for which other dumpsters may be present on the construction site, as shown in Figure 1 . Said types of construction waste may for example include wood, metal, plaster and rubble, or any other type of construction waste present on a given construction site. The binary presence results 16a, 16b are represented as a V for a presence detection or ‘x’ for no detection of a sorting error. Based on the image data of Figure 3a, the binary presence results 16a of the plurality of binary classifiers only show a single positive presence result, in particular for the triangle, so the multiplexer may not output anything or may be configured to output a positive result, i.e. that no construction waste sorting error has been detected. This positive result may, but need not, be signalled to a user, for example by a sign which is generally recognized as positive, as for example a thumbs up 17 or a green light or any other known signal. In contrast, based on the image data of Figure 3b, the binary presence results 16b of the plurality of binary classifiers do show two positive presence results, in particular for the triangle and for the cylinder, so the multiplexer is configured to output a detection of a construction waste sorting error, which may be perceived as a negative result. The method may then trigger a warning device 6 to output a visual and / or acoustic warning signal. The method may further include the step of storing the obtained image data when detecting a construction waste sorting error, such as in the situation of Figure 3b. In this way, users may for example check who has wrongly sorted the construction waste, which may be important information on a construction site where different companies may be working simultaneously.

[0033]

[0029] Note that dumpster 2a of Figure 3a does include a sorting error, in particular, a banana peel 9, but this cannot be detected by one of the binary classifiers 13a, 13b, 13c, 13d since none of them is trained to check the presence of a banana peel 9. To avoid a mixture of construction waste and other types of waste that may be present on a construction site, such as food rests, bottles or cans or the like, the method may further comprise the step of having a further classifier (not shown) check a presence of a further type of waste which is different from the first type and the at least second type, thereby obtaining a further binary presence result. Such a further classifier need not be a binary classifier but may for example be a general classifier trained to recognize a limited number of known types of waste, such as food rests, cans and bottles, or other types of waste. When the outcome of said general classifier is again a binary presence result, the data selector 14 can take said binary presence result into account. Alternatively, said further classifier may also be a binary classifier.

[0034]

[0030] Figures 4a and 4b show schematic views of another preferred embodiment of the method of Figure 2. In particular, Figures 4a and 4b illustrate feedback loops for continuous training of the plurality of binary classifiers. Said plurality of binary classifiers 13a - 13d can be trained in two ways. On every new construction site, or at the beginning of a new phase of construction on given construction site, a new material may pop up as construction waste, and the plurality of binary classifiers may need to be trained or retrained to check a presence of said a new material in the obtained image data. The training may be based on labelled image data representing construction waste of said construction site. Since the classifiers are binary classifiers, such a training may be relatively short. On top of said initial training, the method may comprise steps to provide continuous training to the plurality of classifiers. Figure 4a schematically shows an erroneous output of a detection of a construction waste error. In particular, the binary classifier trained to check a presence of a predetermined construction waste, represented as an arch, may have misinterpreted the banana peel 9 as an arch, resulting in a positive presence result of construction waste of the first type 3a and of the type represented as an arch. Since there are two binary presence results, the method triggers the warning device 6 to output a signal. A user may then check the results, either on the image data obtained from the camera 4, or by checking the content of the dumpster 2a. In this way, the user may recognize the mistake made by the method, conclude that no construction waste sorting error has been made by the method and provide this feedback to the system 1 , for example via interfacing means 7, for example binary feedback buttons, including a first button 18a configured to indicate that no construction waste sorting error has been manually detected, and a second button 18b configured to indicate that a construction waste sorting error has been manually detected in spite of the method not having detected said error, as shown in Figure 4b. Said interfacing means 7 may transmit said binary feedback directly to the controller 5 to retrain the plurality of binary classifiers. In this case, it may not be clear which classifier has made a mistake, so additional information may be provided by the user to the system 1 , for example via a dedicated mobile application, or via additional image data made by a mobile device, as shown in Figure 1. The method may thus comprise the step of receiving binary feedback when a detection of a construction waste sorting error is erroneously output and retraining the first binary classifier and the at least second binary classifier based on said binary feedback, as shown in Figure 4a. The method may further comprise the step of receiving binary feedback when a non-detection of a construction waste sorting error is erroneously output and retraining the first binary classifier and the at least second binary classifier based on said binary feedback, as shown in Figure 4b. The dedicated classifier may not have detected the presence of construction waste of type 3c, as represented by a cylinder, while it is present in the dumpster 2a of Figure 4b configured to receive construction waste of a first type 3a. The method may not have given any output or may have provided a positive output 17. However, a user may have detected the construction waste sorting error and may provide binary feedback to the system via the interfacing means 7, for example by pushing the button 18b. Said feedback may be transmitted to the controller and may provide additional training material, in particular for the binary classifier which has made the error of not detecting construction waste of type 3c.

[0035]

[0031] Figure 5 shows a suitable computing system 500 comprising circuitry enabling the performance of steps of embodiments of the method for detecting a construction waste sorting error according to an aspect of the invention. Computing system 500 may in general be formed as a suitable general-purpose computer and comprise a bus 510, a processor 502, a local memory 504, one or more optional input interfaces 514, one or more optional output interfaces 516, a communication interface 512, a storage element interface 506, and one or more storage elements 508. Bus 510 may comprise one or more conductors that permit communication among the components of the computing system 500. Processor 502 may include any type of conventional processor or microprocessor that interprets and executes programming instructions. Local memory 504 may include a random-access memory (RAM) or another type of dynamic storage device that stores information and instructions for execution by processor 502 and / or a read only memory (ROM) or another type of static storage device that stores static information and instructions for use by processor 502. Input interface 514 may comprise one or more conventional mechanisms that permit an operator or user to input information to the computing device 500, such as a keyboard 520, a mouse 530, a pen, voice recognition and / or biometric mechanisms, a camera, etc. Output interface 516 may comprise one or more conventional mechanisms that output information to the operator or user, such as a display 540, etc. Communication interface 512 may comprise any transceiver-like mechanism such as for example one or more Ethernet interfaces that enables computing system 500 to communicate with other devices and / or systems, for example with other computing devices 581 , 582, 583. The communication interface 512 of computing system 500 may be connected to such another computing system by means of a local area network (LAN) or a wide area network (WAN) such as for example the internet. Storage element interface 506 may comprise a storage interface such as for example a Serial Advanced Technology Attachment (SATA) interface or a Small Computer System Interface (SCSI) for connecting bus 510 to one or more storage elements 508, such as one or more local disks, for example SATA disk drives, and control the reading and writing of data to and / or from these storage elements 508. Although the storage element(s) 508 above is / are described as a local disk, in general any other suitable computer-readable media such as a removable magnetic disk, optical storage media such as a CD or DVD, - ROM disk, solid state drives, flash memory cards, ... could be used.

[0036]

[0032] As used in this application, the term "circuitry" may refer to one or more or all of the following:

[0037] (a) hardware-only circuit implementations such as implementations in only analogue and / or digital circuitry and

[0038] (b) combinations of hardware circuits and software, such as (as applicable):

[0039] (i) a combination of analogue and / or digital hardware circuit(s) with software / firmware and

[0040] (ii) any portions of hardware processor(s) with software (including digital signal processor(s)), software, and memory(ies) that work together to cause an apparatus, such as a mobile phone or server, to perform various functions) and

[0041] (c) hardware circuit(s) and / or processor(s), such as microprocessor(s) or a portion of a microprocessor(s), that requires software (e.g. firmware) for operation, but the software may not be present when it is not needed for operation.

[0042] This definition of circuitry applies to all uses of this term in this application, including in any claims. As a further example, as used in this application, the term circuitry also covers an implementation of merely a hardware circuit or processor (or multiple processors) or portion of a hardware circuit or processor and its (or their) accompanying software and / or firmware. The term circuitry also covers, for example and if applicable to the particular claim element, a baseband integrated circuit or processor integrated circuit for a mobile device or a similar integrated circuit in a server, a cellular network device, or other computing or network device.

[0043]

[0033] Although the present invention has been illustrated by reference to specific embodiments, it will be apparent to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied with various changes and modifications without departing from the scope thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. In other words, it is contemplated to cover any and all modifications, variations or equivalents that fall within the scope of the basic underlying principles and whose essential attributes are claimed in this patent application. It will furthermore be understood by the reader of this patent application that the words "comprising" or "comprise" do not exclude other elements or steps, that the words "a" or "an" do not exclude a plurality, and that a single element, such as a computer system, a processor, or another integrated unit may fulfil the functions of several means recited in the claims. Any reference signs in the claims shall not be construed as limiting the respective claims concerned. The terms "first", "second", third", "a", "b", "c", and the like, when used in the description or in the claims are introduced to distinguish between similar elements or steps and are not necessarily describing a sequential or chronological order. Similarly, the terms "top", "bottom", "over", "under", and the like are introduced for descriptive purposes and not necessarily to denote relative positions. It is to be understood that the terms so used are interchangeable under appropriate circumstances and embodiments of the invention are capable of operating according to the present invention in other sequences, or in orientations different from the one(s) described or illustrated above.

Claims

CLAIMS1. Computer-implemented method for detecting a construction waste sorting error on a construction site including at least one dumpster for receiving construction waste, the method comprising the steps of- obtaining image data representing a content of a first dumpster for receiving construction waste of a first type;- checking by a first binary classifier a presence of said first type of construction waste in the content of the first dumpster, thereby obtaining a first binary presence result;- checking by at least a second binary classifier a presence of at least a second type of construction waste in the content of the first dumpster, thereby obtaining an at least second binary presence result;- outputting a detection of a construction waste sorting error when said first binary presence result and said at least second binary presence result include at least two positive presence results, characterized in that said first binary classifier and said second binary classifier are individually trained exclusively on detecting one type of construction waste.

2. The method according to claim 1 , further comprising the step of receiving binary feedback when a detection of a construction waste sorting error is erroneously output and retraining the first binary classifier and the at least second binary classifier based on said binary feedback.

3. The method according to any of the preceding claims, further comprising the step of receiving binary feedback when a non-detection of a construction waste sorting error is erroneously output and retraining the first binary classifier and the at least second binary classifier based on said binary feedback.

4. The method according to any of the preceding claims, further comprising the step of training said first binary classifier and the at least second binaryclassifier using labelled image data representing construction waste of said construction site.

5. The method according to any of the preceding claims, further comprising the step of checking by a further classifier a presence of a further type of construction waste which is different from the first type and the at least second type, thereby obtaining a further binary presence result.

6. The method according to any of the preceding claims, further comprising the step of storing the obtained image data when detecting a construction waste sorting error.

7. The method according to any of the preceding claims, wherein the obtained image data further represent a content of at least a second dumpster for receiving construction waste of at least a second type and wherein the method further comprises a step of selecting a single dumpster among the first dumpster and the at least second dumpster represented in the image data.

8. The method according to any of the preceding claims, wherein said first type and said at least second type of construction waste includes one of wood, metal, plaster, building rubble.

9. The method according to any of the preceding claims, wherein the obtained image data is a greyscale image.

10. A controller comprising at least one processor and at least one memory including computer program code, the at least one memory and computer program code configured to, with the at least one processor, cause the controller to perform the method according to any of the preceding claims 1 - 9.11 .A computer program product comprising computer-executable instructions for performing the method according to any of the preceding claims 1 - 9 when the program is run on a computer.

12. A computer readable storage medium comprising computer-executable instructions for performing the methods according to any of the preceding claims 1 - 9 when the program is run on a computer.

13. System for detecting a construction waste sorting error, the system comprising:- at least one camera configured to capture image data of a content of at least one dumpster for receiving construction waste of at least a first type; - a controller according to claim 10.

14. System according to claim 13, the system further comprising a warning device configured to visually and / or acoustically output a warning signal when triggered by the controller.

15. System according to any of the preceding claims 13 - 14, the system further comprising interfacing means configured to allow a user to provide feedback.

Images