Research Output
1 - 4 of 4
-
PublicationFeasibility Study of Beef Quality Assessment using Computer Vision and Deep Neural Network (DNN) Algorithm( 2020-08-24)The beef quality relies upon the colour score of muscle during the grading stage. Colour scoring to be used in beef grading would be very critical and the current way of identification and determination of the quality of beef is still being done manually and susceptible to human error. The ability to automate the prediction of the beef quality will assist the meat industry through the grading phase to establish the colour score. Therefore, computer vision and deep neural network (DNN) were used to predict the beef quality by determining colour scores of beef muscle. Four hundred of beef rib-eye steaks were chosen and acquired for each image, which is the colour score of beef were assigned by expertise according to the standard colour cards. The image was processed and went through DNN classifier for determining beef quality. The proposed DNN classifier achieved the best performance percentage of 90.0%, showing that the computer vision integrated with the DNN algorithm can deliver an efficient implementation for predicting beef quality using colour scores of beef muscle.
1 -
PublicationDevelopment of automated hybrid intelligent system for herbs plant classification and early herbs plant disease detection( 2020-08-01)Plants such as herbs are widely used in the medical and cosmetic industry. Recognizing a species and detecting an early disease of a plant are quite challenging and difficult to implement as an automated device. The manual identification process is a lengthy process and requires a prior understanding about the plant itself, such as shape, odour, and texture. Thus, this research aimed to realize the computerized method to recognize the species and detect early disease of the herbs by referring to these characteristics. This research has been developed a system for recognizing the species and detecting the early disease of the herbs using computer vision and electronic nose, which focus on odour, shape, colour and texture extraction of herb leaves, together with a hybrid intelligent system that are involved fuzzy inference system, naïve Bayes (NB), probabilistic neural network (PNN) and support vector machine (SVM) classifier. These techniques were used to perform a convenient and effective herb species recognition and early disease detection on ten different herb species samples. The species recognition accuracy rate among ten different species using computer vision and electronic nose is archived 97% and 96%, respectively, in SVM, 98% and 98%, respectively, in PNN and both 94% in NB. In the early disease detection, the detection rate among ten different herb’s species using computer vision and electronic nose are 98% and 97%, respectively, in SVM, both 98% in PNN, 95% and 94%, respectively, in NB. Integrated three machine learning approaches have successfully achieved almost 99% for recognition and detection rate.
1 -
PublicationDesign and Development of GreedLea Routing Protocol for Internet of Vehicle (IoV)( 2020-03-20)In Internet of Vehicle (IoV), each vehicle uses a routing protocol to find a path for sending its messages to the last destination. Nowadays, the studies of IoV routing protocols and their impact on the performances of network with different network scenarios has significantly developed a precise understanding of the requirements and goals for designing an IoV routing protocol. In IoV, topology of network diverse promptly which leads to the fragmentation of network, frequent route breakage, and packet loss. This paper discusses on the development of an integrated routing protocol for IoV scenario. Greedy Perimeter Stateless Routing (GPSR) and Reinforcement Learning (RL) is integrate to determine a route based on demand. Then, the mobility model has been designed to reduce road collision. Lastly, traffic management also been focused to deal with the loss, mobility and network delay to meet the application demands.
1 -
PublicationAutomated tomato grading system using Computer Vision (CV) and Deep Neural Network (DNN) Algorithm( 2022-01-01)The tomato grading is based on the skin colour at the grading stage. The evaluation of the colour used to classify tomatoes is very important, and the current methods of identifying and determining tomato varieties are still manual and prone to human error. The ability to automate tomato grading helps the food industry determine colour grades during the evaluation phase. Therefore, Computer Vision (CV) and Deep Neural Network (DNN) are utilised to grade tomatoes by determining their maturity colour. Three hundred tomatoes were selected and its maturity level are assigned by expertise. The tomato images are captured, processed and passed to the DNN classifier to determine the tomato grade. The proposed DNN classifier achieved the mAP percentage of 95.52%. This shows that the computer vision built into the DNN algorithm can provide an efficient implementation for predicting tomato grade.
3 1