Image Segmentation and Maturity Recognition Algorithm Analysis Report

This report analyzes an image segmentation and maturity recognition algorithm proposed for identifying Lingwu long jujubes. The study examines the algorithm's methodology, focusing on color difference fusion in RGB color space to segment images and recognize fruit maturity. The report details the parameters used, results obtained, and the algorithm's effectiveness compared to previous methods, including the use of L*a*b* color space. It highlights the algorithm's strengths, such as its robustness against light variations and its efficiency in maturity recognition. The report also discusses the weaknesses, including limitations in handling non-red areas and the impact of occlusion. Furthermore, it suggests avenues for future research, such as expanding the algorithm to handle a larger number of objects and addressing adhesion and occlusion phenomena. The report concludes with a comprehensive overview of the algorithm's performance and potential for improvement.

Running head: IMAGE PROCESSING
IMAGE PROCESSING
Name of the Student
Name of the University
Author Note

1
IMAGE PROCESSING
Abstract:
The aim of the report is to identify the image segmentation method as identified in the
provided journal. The report has particularly looked for the method in details. This details
include the effectiveness of the proposed algorithm and the improvement it brings compared
to other algorithms proposed previously by other researchers. Along with the application of
the algorithm proposed by the author, the issues and weakness of the paper has also been
discussed. The scope of further research has also been described.

2
IMAGE PROCESSING
Table of Contents
Introduction to the topic:............................................................................................................2
Key findings of the previous authors:........................................................................................2
Methodology adopted in given paper.........................................................................................2
Parameters used for analysis......................................................................................................3
Results obtained.........................................................................................................................3
Weaknesses in the paper............................................................................................................4
Scope of further investigations...................................................................................................4
References:.................................................................................................................................5

3
IMAGE PROCESSING
Introduction to the topic:
Image segmentation plays a key role in the fruit recognition and hence it is important
is to have an algorithm that is robust and provide an excellent means for image segmentation
as well (Sowmya and B.S. 2015). In order to improve the efficiency of the automatic fruit
picking robots an image segmentation method has been studied (Xiong et al. 2015). Along
with that the watershed transformation method has been studied that is based on gradient and
distance.
Key findings of the previous authors:
 Segmentation algorithm mostly designed on the basis of color threshold and hence
there is limitation in those algorithms due to circumstances and light conditions which
affect the performance of the algorithm and the quality of the image segmentation and
therefore the image recognition.
 occlusion and adhesion phenomena also decrease the quality of the segmentation
algorithms
Methodology adopted in given paper
In order to study the efficiency of the algorithm proposed here, Lingwu long jujubes’
images were captured with digital camera FUJIFILM. Here the author have went for
quantitative research method which has allowed the author to assess the effectiveness of the
proposed algorithm without any bias and perform an original research on the topic (Dan
2015). The experimental result has been published after the algorithm has been applied on the
samples taken with the digital camera. 997 pieces out of 1665 taken photos were considered
in different circumstances such as touching, covering and shadow. 50 pictures were further
shortlisted where the number of jujubes were less than. The experimental result has provided
the research empirical validation as well.

4
IMAGE PROCESSING
Parameters used for analysis
The parameters that have used in this papers are L, a, b which is related to the L* a*
b* space and each of these three parameters corresponds to the three colour component of the
L* a* b* (Chan 2014). The RGB colour space is possible to convert into several other colour
spaces such as HIS and HSV space, and YCbCr (Zhou et al. 2017). While each of the colour
and each of the associate parameters have pros and cons, L* a*b* space is relatively robust
against variations of light intensity and it is most uniform colour space as well (Domingo et
al. 2014). However it is not possible to directly transformed RGB to L* a* b*, XYZ space is
considered as the intermediate transform with parameters X, Y, Z associated with XYZ space
(Warren 2017).
Results obtained
The experiment result has been presented in two separate section. The first section or
the table 1 has described the result of the segmentation testing of 100 images and the
experiment has been conducted with natural light. As per the data included in the table the
mean rate of correct segmentation (ration of correct segmentation to total segmentation that
includes both correct and wrong segmentation) is 93.27%.
In the next section comparison between maturity recognition of Lingwu long jujubes
and manual recognition has been described. The error rate of maturity recognition is 10%
while the recognition criteria is also higher than the manual recognition.
Important conclusions derived based on the result
This method is based on the colour difference fusion in RGB colour space. This is an
excellent technique for performing the image segmentation (Bocken and Short 2016). It is
more powerful and efficient than the manual recognition (Laukkanen 2015). It helps to

5
IMAGE PROCESSING
implement the maturity recognition which is effective for image recognition in complex
environment (Reijonen et al. 2015).
Weaknesses in the paper
 The method that has been adopted in this paper is based on correct segmentation of
only red areas of the jujubes has been considered. Non-red areas are affected by peel
shadows.
 The correct rate that is associated with the maturity recognition is depends on the
threshold values and if the threshold value is not proper segmentation will not be
correct (Porterfield 2015)
 Although The error rate provided by the algorithm is excellent ,it increases with the
increase in the occlusion area and number of jujubes (Raisch 2016)
 The Number of jujubes selected for applying the algorithm has been limited to 8 and
also the image has been taken in natural scene for accurate result.
Scope of further investigations
There is scope for further research in this context. In order to increase the efficiency
of the algorithm it should consider to execute the image segmentation process with more than
8 objects, for example at least 10. In case of long jujubes’ image, severe adhesion and
occlusion phenomena should also be considered which offers an excellent opportunity to
conduct further research in this topic.

6
IMAGE PROCESSING
References:
Bocken, N.M.P. and Short, S.W., 2016. The theory of image segmentation in fruit
recognition: a comprehensive discussion. McGraw-Hill, Inc.
Chan, C., 2014, January. The importance of empirical validation in research.
Dan, A., the scope and status of image segmentation in fruit recognition, 2015. Method and
apparatus for leading effective algorithm design.
Domingo, D.L., B.P. Serrano, E.P. Serrano and E.J. del Rosario, 2014. Digital photometric
method for determining degree of harvest maturity and ripeness of 'sinta' papaya (Carica
papaya L.) fruits. Philipp. Agric. Sci., 95(3): 252-259.
Laukkanen, T., 2015. The definition and scope of image segmentation in object recognition.
A modern theory of image processing, 42, pp.35-46.
Porterfield, T.E., 2015. Evaluation of image segmentation: an empirical investigation of
scope and success. International Journal of information and technology, 40(6), pp.435-455.
Raisch, W., 2016. Towards a sufficiency-driven image recognition: evaluation of as-Is
Workflow Modelling, 18, pp.41-61.
Reijonen, H., Hiroden, S., Nagy, G., Laukkanen, T. and Gabrielsson, M., 2015. The bias of
research method in research. A modern approach, 51, pp.35-46.
Sowmya, B. and B.S. Rani, 2015. Colour image segmentation using fuzzy clustering
techniques and competitive neural network. Appl. Soft Comput. 11(3): 3170-3178.
Warren. N, 2017, January. The importance of post evaluation for successful research
implementation: an intuitive guide.

7
IMAGE PROCESSING
Xiong, J.T., X.J. Zou, L.J. Chen and A.X. Guo, 2015. Recognition of mature litchi in natural
environment based on machine vision. Trans. Chin. Soc. Agric. Mach., 42(9): 162-166.
Zhou, T.J., T.Z.H. Zhang, L. Yang and J.Y. Zhao, 2017. Comparison of two algorithms based
on mathematical morphology for segmentation of touching strawberry fruits. Trans. Chin.
Soc. Agric. Eng., 23(9): 164-168.