Modified Deep-Convolution Neural Network Model for Flower Images Segmentation and Predictions

Nowadays, recognition of plant, leaf, and flower images is one of the most challenging issues due to the wide variety of classes on earth, which are based on amount of texture, color distinctiveness, shape distinctiveness, and different sizes. This paper proposes a hybrid method, known as Modified Deep-Convolution Neural Network Model (MDCNN) for the segmentation and recognition of flower images that employs Deep Convolution Neural Network with a combination of color model and image processing. Initially, L*a*b color space conversion is applied to reduce the multi-dimensions and geometry of images in which the red-green axis, blue-yellow axis, and luminosity are represented by chromaticity layers a*, b* and L* respectively. Moreover, the model also takes an input of different flower images and converts the RGB color model into the L*a*b color model while reducing the effort of image segmentation. That is performed by the canny edge detection algorithm. Moreover, a deep convolutional neural network with hidden layer is designed for classification and prediction of flowers with five different classes like daisy, dandelion, rose, tulip, and sunflower. This paper also represents the minimum computation time of MDCNN to detect flowers, along with the CPU and GPU. It is also compared with pre-trained convolutional neural networks such as VggNet-16, GoogleNet, AlexNet, and ResNet-50 in terms of f1-score, accuracy, precision, and sensitivity. Finally, the proposed method accurately recognised flower images with accuracy up to 98%, maximising up to + 1.89% from the state of the art while minimising the image segmentation error rate.