Edge detection using neural network arbitration

Ramalho, Mário

doi:10.1049/cp:19950712

Cited by 7 publications

(4 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Once the network converges (learns), this phase will only comprise running the trained neural network using one forward pass, and the result would be identifying the chicken portion. Neural network arbitration of images and edge maps has been successfully used in many applications over the past two decades (Ramalho and Curtis 1994, 1995).…”

Section: Methodsmentioning

confidence: 99%

Automatic Identification System for Raw Poultry Portions

Khashman¹

2011

J Food Process Engineering

View full text Add to dashboard Cite

In poultry processing plants, the sorting of raw poultry portions into separate containers prior to packaging is performed mostly by human laborers. This sorting method has two problems: labor cost and health hazard due to possible contamination between raw meat and humans. Therefore, an automated sorting system is required to avoid these problems. In this article, we present an automatic portion identification system that uses image processing and a neural network to identify six different chicken portions that are usually preferred by consumers. The proposed rotational invariant system can identify the cut portions regardless of their orientation with an overall correct identification rate of 97.57%. Potentially, the output of the proposed identification system can be used to move robotic arms to physically separate the identified chicken portions into separate containers. PRACTICAL APPLICATIONS This work contributes to the development of an automated sorting system for practical use in poultry processing plants. The presented intelligent identification system is the main part of such an automated system. We train the system in this work to recognize the main chicken portions; however, other portions and other poultry birds can also be identified upon including their portion images in the training phase. This work differs from existing systems, in that it relies totally on the shape and coarse texture of a portion, using images as input data, and discards information like weight or size of the portion. Another advantage is the elimination of the need for using many birds during the development of the system; in fact, training the system can be achieved using only one bird and its portions.

show abstract

Section: Methodsmentioning

confidence: 99%

Automatic Identification System for Raw Poultry Portions

Khashman¹

2011

J Food Process Engineering

View full text Add to dashboard Cite

show abstract

“…As a result, in order to acquire accurate edge, threshold T is replaced by the valleyemphasis threshold * T [9] to become the optimal threshold. The optimal threshold * T can be defined by maximizing the between-class variance; that is: Fig 2, it can be found that the region of the fuzzy enhancement pixels is [ 1,2] h h , when linear mapping transformation is selected as the fuzzy membership. However, a convex curve mapping transformation will make the region become [ 1,2] f f which is larger than the region of [ 1,2] h h .…”

Section: IIImentioning

confidence: 99%

“…However, most edge detectors are sensitive to noise, including the conventional methods such as Sobel, Prewitt and so on. About this issue, many approaches of edge detection based on neural network [1], genetic algorithm [2], and wavelet theory [3] have been presented. In addition, due to the fuzziness of noise image edge, many authors adopt fuzzy reasoning in order to extract edge.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Improved Edge Detection in Noisy Image Using Fuzzy Enhancement

Chen

2010

2010 International Conference on Biomedical Engineering and Computer Science

View full text Add to dashboard Cite

To realize the edge detection of the weak target objects comparing with the other objects in noisy image, an improved edge detection algorithm using fuzzy enhancement is presented. The algorithm is developed through analyzing the defects of the fast multilevel fuzzy edge detection algorithm. It not only achieves optimal threshold in membership function, but also proposes the improved membership function to make the region of the fuzzy enhancement pixels larger. In addition, processing the median filter of fuzzy sets is used to reduce pulse noise, and fuzzy entropy operator is adopted to detect edge fast. The result shows the proposed algorithm is reasonable and effective.

show abstract