Neural network application to container number recognition

Lui, Ho Chung; Lee, C.M.; Gao, Fuqing

doi:10.1109/cmpsac.1990.139350

Cited by 8 publications

(4 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is necessary to extract and recognize container identity automatically by image process and pattern recognition techniques. Though automatic container code recognition had been studied since 1990's [1,2], there were still many difficulties in container code recognition, especially in code detection and char extraction. The main difficulties come from: (1) Container code characters were painted with a wide range of sizes, fonts, colors, and spacing.…”

Section: Introductionmentioning

confidence: 99%

Container Code Detection by Deep Convolutional Network

Wang¹,

Ma²

2017

Proceedings of the 2017 Global Conference on Mechanics and Civil Engineering (GCMCE 2017)

View full text Add to dashboard Cite

Abstract. Automatic container code recognition plays an important role in customs logistics and tra-nsport management. Due to difficulties such as character color and font size variation, illumination conditions, image degradation, and exist of many other characters, automatic detect and recognition of container code is still a difficult task. This paper proposes a container code detection algorithm based on deep convolutional neural network named holistically-nested edge detection (HED). In the training phase, a bounding box was drawn around container code as virtual edges, and they were feed to the network together with original image to train the HED model. In the test phase, probability map of bounding box was predict by trained model and finally bounding box is obtained by thresholding and connected region analysis. Experimental results on 9953 container images show that the performance of IOU and recall precision on test set can reach 0.646 and 0.934 respectively with the proposed algorithms.

show abstract

Section: Introductionmentioning

confidence: 99%

Container Code Detection by Deep Convolutional Network

Wang¹,

Ma²

2017

Proceedings of the 2017 Global Conference on Mechanics and Civil Engineering (GCMCE 2017)

View full text Add to dashboard Cite

show abstract

“…The neural network in the VECON system, which has been applied in a similar problem [4], acts as the character recognition module. Basically, it is a general feed-forward multi-layer perceptron with one hidden layer, trained by the back-propagation algorithm Given an input character pattern, the score of each output node resulted from the network operation describes the likeliness for the input to be regarded as the category associated with that node.…”

Section: Neural Network Character Recognizermentioning

confidence: 99%

“…1 A three-line container number processing (identifying) truck and container ID numbers, a situation that has attracted our attention because of the potential of automating such a process. Many methods [1]- [4] have been proposed for character recognition but they are often subjected to substantial constraints. We have developed a practical vision system for vehicle and container ID numbers recognition called VECON.…”

Section: Introductionmentioning

confidence: 99%

Automatic character recognition for moving and stationary vehicles and containers in real-life images

Lee

IJCNN'99. International Joint Conference on Neural Networks. Proceedings (Cat. No.99CH36339)

View full text Add to dashboard Cite

Many methods have been proposed for character recognition but they are ojlen subjected to substantial constraints, due to unexpected d@culties encountered in real-li&e images. A real-lge image may be complex for a varieiy of reasons. Rust, m d , peeling paint, or fading color may distort the images of the characters: uneven lighting may make them di9cult to discern. This paper presents the ECON (vehicle and Container Number Recognition) system, which takes into account a wide range of real-life consi&rations and aims at oflering applicable solutions to some industries. Ajler being tested under outdoor environment and 24-hour operations, the proposed methods proved to have accuracy higher than 95%. f i e system was commercially employed in car parks, bus stations, border checkpoints, container terminals and container depots.

show abstract

“…These weights are also involved in the S-cell weight update process during supervised training (Algo- From the time that the neocognitron was first proposed, there appears to have been some level of confusion among experimenters as to how to calculate the masks for the S and C-cells. Some authors have chosen to omit reference to these weights altogether [93,118,132,165,183,184] others invent masks of their own [124,168]; only one person has suggested their irrelevance [77]. In an attempt to clarify the matter, two methods for calculating ct{i/) and d({i/) will be reviewed.…”

Section: The Mask Parameters and Their Effect On The Neocognitronmentioning

confidence: 99%

The neocognitron as a system for handwritten character recognition : limitations and improvements

Lovell¹

View full text Add to dashboard Cite

This thesis is about the neocognitron, a neural network that was proposed by Fukushima in 1979. Inspired by Hubel and Wiesel's serial model of processing in the visual cortex, the neocognitron was initially intended as a self-organizing model of vision, however, we are concerned with the supervised version of the network, put forward by Fukushima in 1983. Through "training with a teacher", Fukushima hoped to obtain a character recognition system that was tolerant of shifts and deformations in input images. Until now though, it has not been clear whether Fukushima's approach has resulted in a network that can rival the performance of other recognition systems.In the first three chapters of this thesis, the biological basis, operational principles SHOP serves as a method for probing the behaviour of the neocognitron and is used to investigate the effect of cell masks, skeletonization of input data and choice of training patterns on the network*s performance. Even though SHOP is the best selectivity adjustment algorithm to be described to date, the system's peak correct recognition rate (for isolated ZIP code digits from the CEDAR database) is around 75% (with 75% reliability) after SHOP training. It is clear that the neocognitron, as originally described by Fukushima, is unable to match the performance of today's most accurate digit recognition systems which typically achieve 90% correct recognition with near 100% reliability.After observing the neocognitron's failure to exploit the distinguishing features of different kinds of digits in its classification of images, Chapter 6 proposes modifications to enhance the networks ability in this regard. Using this new architecture, a correct clcissification rate of 84.62% (with 96.36% reliability) was obtained on CEDAR ZIP codes, a substantial improvement but still a level of performance that is somewhat less than state-of-the-art recognition rates. Chapter 6 concludes with a critical review of the hierarchical feature extraction paradigm.The final chapter summarizes the material presented in this thesis and draws the significant findings together in a series of conclusions. In addition to the investigation of the neocognitron, this thesis also contains a derivation of statistical bounds on the errors that arise in multilayer feedforward networks as a result of weight perturbation (Appendix E).

show abstract

Neural network application to container number recognition

Cited by 8 publications

References 6 publications

Container Code Detection by Deep Convolutional Network

Container Code Detection by Deep Convolutional Network

Automatic character recognition for moving and stationary vehicles and containers in real-life images

The neocognitron as a system for handwritten character recognition : limitations and improvements

Contact Info

Product

Resources

About