GSA Object Counter
recognition and classification of objects
Software to automatic recognition and classification of two-dimensional objects. The product consists of two analysis methods for object recognition, which are independent of each other. Artificial intelligence (AI) and Color analysis is used.
The GSA Object Counter is a software developed for automatic recognition and classification of two-dimensional objects. The product consists of two analysis methods for object recognition, which are independent of each other.
Method number one uses a neural network. Artificial intelligence (AI) is used to try to assign new objects to trained object groups. This procedure was tested in a practical approach with different objects (objects, numbers) and produced very good results.
The second object recognition method is a color-based analysis method. Also this method has proven itself in practice (detection of different PET bottles). The method, using a neural network, is suitable for objects that differ from one another in terms of their object shape. The color analysis method is to be preferred if objects have the same shape but differ in their coloring.
Method number one uses a neural network. Artificial intelligence (AI) is used to try to assign new objects to trained object groups. This procedure was tested in a practical approach with different objects (objects, numbers) and produced very good results.
The second object recognition method is a color-based analysis method. Also this method has proven itself in practice (detection of different PET bottles). The method, using a neural network, is suitable for objects that differ from one another in terms of their object shape. The color analysis method is to be preferred if objects have the same shape but differ in their coloring.
Color Analysis
Settings for the color analysis method. The criteria for creating color groups are configured here. These color groups are later used for object recognition in the color analysis process.
Neural Network
Configuration window for the neural network. The dimension and the fault tolerance of the network are configured here. Image filters that support object recognition can also be defined.
General Settings
Object limitation via object size. Limiting the minimum and maximum number of pixels per object helps to distinguish between the object and its surroundings.
Basic settings and environment
Setting analysis method and image release criteria. Here you can choose between the two available analysis methods. The tolerances for triggering the image acquisition can be configured here.
Timer-based image monitoring
The start and end of monitoring can be set using a timer. Define when to start or stop the automated processing.
Monitoring of defined image areas
The monitored image area can be restricted. In this way, errors in object recognition can be avoided.
AI in Action
The analysis is complete. All images were sorted into the correct object classes. The correspondence with the respective object group is output as a percentage.
Training Group Settings
All training images of an object group are displayed for control purposes. It is possible to remove images from the group.
Configurable Result Output
The form and scope of the result output is fully adjustable. In this way, the results of the analysis can be configured for further processing steps. Integration into other programs is therefore given.
Training Room
Assignment of the training images to the corresponding object groups for the color analysis process. This manual assignment is necessary once. After the training, the object is recognized automatically.
