Case Study: Food Logistics „At its best“ – Automation with Vision Technology

Case Study: Food Logistics „At its best“ – Automation with Vision Technology

Price competition in the food trade is extremely tough. This is why efficient processes play an important role, and logistics is no exception. FRAMOS worked with a major food retailer to automate its central warehouse. FRAMOS solutions were installed at various processing stations, using traditional imaging processing, self-learning algorithms and classifiers to ensure robust and smooth operation. For example this enables the systems in the ‘empty containers centre’, to recognise different container types and select the correct process path. This recognition is possible even if they are covered in dirt or partially occluded by large labels.

A supermarket chain in Switzerland is currently using FRAMOS expertise, ‘FRAMOS Sorting Intelligence’ along-side its own system configurations to perform four different sorting and recognition tasks. These are part of the delivery chain within its logistics centre. These provide an insight into the benefits Imaging can provide to all retail and mail order sectors as optimising logistics is essential to lower costs, increase throughput and increase reliability.

1) Fresh food delivery: Identifying container types

Typically, fresh goods from various producers are delivered to the logistics centre in containers of varying shape, size and colour. In fresh produce logistics, a distribution centre is a cold store, a building cooled down to three degrees Celsius. When the goods arrive, they are taken into the cold store by conveyor. These containers are then stacked onto trolleys for transport within the cold store. For stability and to prevent damage to containers and their contents. Only containers of the same type can form a layer in a stack, however the layers can made from any container type.

The container tower is around two metres high, and is destacked using a wedge conveyor on the upper end of a lifting device. To ensure that this process functions correctly, the type of the container at the top of the stack and that below it must be recognised. This is where the first of the FRAMOS systems comes in. A dark field illumination system, created with 6 LED bar lights from Falcon is used to illuminate the containers. This lighting scenario highlights the grip recesses and perforations in the sides of the containers. This is then imaged using a 1.3 Mpixel camera from Smartek Vision, the GC1391M. The image acquired is of the top two layers of the container tower, and is passed to the machine learning classifier. Based on the previously “learned” images, this classifier can identify the containers in these layers. The results are then transferred to the system control, and the top layer is de-stacked. The entire container tower is then raised by the height of the next layer, which is now known, and the process begins again until there are no more layers and the tower is completely destacked.

By using learning algorithms, the FRAMOS system is able to correctly classify every container. These are often damaged, dirty or covered with paper labels. The system can currently classify 20 different container types reliably. Another major advantage of the new approach is that it is easy to add new container types to the existing ones. All that is required is to generate images of the new containers in the system and teach them in a “learning phase”. A rules-based classification system would severely limit this possibility.

Benefits: By using learning algorithms instead of rules-based classifiers, the system can reliable classify different container types within a single class, even with a high variety of types. This makes it ideally suited for varied sorting tasks in logistics.

Kiste1   Kiste2

 

2)      Empty containers centre: Automatically unfolding of IFCO containers

The Swiss supermarket chain uses foldable, reusable containers, known as IFCO containers, to transport fruit and vegetables to it’s stores. After the goods are delivered, the containers are folded and returned in stacks. These plastic, standardised containers are available in “half” and “full” sizes, measuring 300 mm and 600 mm in length respectively. Only half containers can be stacked on top of full-size containers and not vice versa. The containers are returned to the ‘empty containers centre’ as a tower of up to 1.80 metres in height. They must be automatically de-stacked before being processed further. The 3-axis robot system used to do this needs to know the precise number of layers of full containers in the lower part of the container stack, and the number of half container layers in the upper part. To achieve this, the FRAMOS measurement system projects a vertical laser line covering the entire height of the stack. The stack profile this creates is then captured by a 5 Mpixel CCD camera (Smartek GC2441) fitted with a with bandpass filter paired with the laser. The specific pattern of the full and half-size containers, along with the transitions, makes it possible to reliably differentiate the container types and count the layers. This information is provided to the robot controller before every gripping process.

This means the robot can build stacks of a single type, which are then automatically conveyed to the appropriate washer. A total of two IFCO destackers are in use in this installation.

Benefits: Using line lasers and bandwidth filters ensures greater reliability in the event of changes in the ambient lighting, as well as variations in the object presentation such as dirt or colour changes. The camera’s high resolution enables reliable evaluation of the laser line for the entire height of the stack.

3) Internal washing: Container sorting before the washer

In addition to the IFCO containers, all other non-foldable container types must be sorted in the empty containers centre. One task here is to separate the company’s own three container types from those of other manufacturers, so that these can be washed before being used again. To achieve this, FRAMOS VLG dimension measurement systems are integrated in a total of eight conveyor lines. These systems consist of a horizontal and vertical light grid. As an object passes through this ‘light curtain’, a part of this grid is blocked, this provides information about the contours of the measured object. All containers on the conveyor pass through the volume light grid at a speed of 0.65 metres per second. The VLG measures the outer dimensions and the hole pattern. Two of the company’s three containers can be differentiated from those of other suppliers purely by their dimensions. The third container type has standard dimensions, however, it can be differentiated from other manufacturers’ containers based on the structure of the side pattern. By efficiently extracting the characteristics of this hole pattern, a descriptor was generated that enables reliable container classification without this being jeopardised by dirt or covered areas.

Each container that passes through the measurement stations sends a result to the system control which operate pushers, which push the company’s own containers from the conveyor.

Benefits: Using the FRAMOS VLG volume measurement system represents a simple and extremely reliable alternative to camera-based type recognition that is easy to maintain. It also minimised the software development workload.

4)      Empty containers centre: High-speed classification and sorting of containers

Once the company’s own containers have been removed, the other containers must be sorted so that they can be sent back to the respective suppliers. There are 16 different container types in total here, not all of which can be differentiated by the outer dimensions. This high-speed system, 1.75 metres per second, features FRAMOS’ fastest light grid system: the VLG SHS (Super High Speed). This can achieve a length resolution of better than 3 mm. In addition to analysing the dimensions and hole pattern, this measurement station also uses a colour camera, as some containers can only be differentiated by colour. FRAMOS image processing is still able to accurately classify the containers with colour recognition even with containers that have seen many years of intensive use and could have dirt and attached labels, even within a single class.

The result of the classification is transmitted to the controller, which allocates the containers to the correct discharge chute. At the end, a robot builds new towers consisting of a single type. Organised by size, the containers are then sent back to the respective supplier.

Benefits: This system successfully uses the same algorithms as the type recognition in the washer. The only additional component required is a reliable colour descriptor.

Stay Tuned with FRAMOS. Sign Up for our Monthly Newsletters

Contact Us

Salutation*
First Name*
Last Name*
Job Title*
Company*
Email*
Phone*
Country*
What FRAMOS product or service would you like to know more about?*
By submitting the request, your data will be processed to get in contact with you and answer your questions. For more information about the processing of your data, please take a look at our  privacy policy.

Contact Us

Salutation*
First Name*
Last Name*
Job Title*
Company*
Email*
Phone*
Country*
By submitting the request, your data will be processed to get in contact with you and answer your questions. For more information about the processing of your data, please take a look at our  privacy policy.