This is because classification using shape signatures is computationally intensive due to the complex normalization needed to obtain rotational invariance. The shape signature will only be rotationally invariant if the starting point on our original shape can be identified on our rotated shape. This may not always be obvious.
In addition, because shape signatures are local representations of shape features extracted from the spatial domain, they are sensitive to noise. Fourier descriptors have the advantages of simple computation and also simple normalization making it preferable for online retrieval. We show later that the Fourier descriptors obtained from the centroid distance function are also robust to noise, yet another advantage.
The centroid distance function expresses the distances of the boundary points from the centroid (gx,gy ) of a shape. It is given by the following formula:
r(t)=√([(x(t)-g_x )^2+(y(t)-g_y )^2 ] ) ..(3.8)
Since this shape signature is only dependent on the location of the centroid and the points on the boundary, it is invariant to the translation of the shape and also the rotation of the shape if the starting point of the original shape can be identified on the rotated shape. If this is possible and we calculate our new shape signature starting at that point, then the shape signatures of the original and rotated shapes will be identical. This signature alone is not invariant to a change in starting point or scaling so we apply the Fourier transformation, normalize the coefficients, and then take the magnitude of these normalized coefficients to obtain a descriptor that is invariant to translation, rotation, scaling, and change of starting point.
Distance from centroid to centroid is calculated. These distances are calculated from each centroid to other remaining centroids, which then considered for average distance of centroids.
Size is a very important feature of the image expression which is generally considered a closed area surrounded by contour curve in the two-dimensional image space. There are four major characteristics values in size-based image feature extraction. (1) Geometry features: Including circumference, area, axis, orientation angle and so on. (2) Area description features: Using a set of characteristics that describe its features based on the target area. (3) Moment invariants: the features describe the geometric characteristics of the region, commonly used Hue invariant moments, orthogonal moments, etc. (4) Fourier size descriptor.
The methods of shape feature extraction have been widely used in image analysis of crop disease. It improves the accuracy and efficiency of disease diagnosis. However, the shape-based image feature extraction methods are suitable to the lesion which shapes are easily identifiable, such as some simple type of graph lines. When the shapes of lesion are more complex, the results of extraction are not very satisfactory. In addition, the shape feature extraction must be based on the area divided of lesion.
The result of image segmentation has a direct impact on the effect of shape feature extraction. We could extract features of crop diseases through color, texture, shape and other characteristics with the crops lesion. However, the color features present in the leaf surface will be different along with the different phases of crop disease and the texture feature extraction is vulnerable to the constraints of the image clarity conditions. The shape features are more stable than the color features and better than the texture feature extraction.
An artificial neural network (ANN), often just called a ‘neural network’ (NN), is a mathematical model or computational model based on biological neural network. It consists of an interconnected group of artificial network and processes information using a connectionist approach to computation.
There are many types of neural network for various applications available in the literature, we need a supervised network training and classification, several network models are available. The most commonly used and simplest network architecture called Feed– Forward Back propagation neural network (FFBPNN). If data has been properly preprocessed, the classification rate of back propagation networks is similar to more sophisticated networks, as Radial Basis Function (RBF) networks. According to the literature survey BPNN gave high accuracy among the different Neural Networks used.
True neural networks are biological systems (BRAIN) that detect patterns, make predictions and learn. The human brain is a very complex part of the human body; due mainly to the interactions and connectivity with other parts of our body, and the way it controls and defines every aspect of our being. The purpose of a neural network is to learn to recognize patterns in data. Once the neural network has been trained on samples, it can make predictions by detecting similar patterns in future data. Neural networks provide a range of powerful new techniques for solving problems in pattern recognition, data analysis, and control.
ANNs have been applied to an increasing number of real-world problems of considerable complexity. Their most important advantage is in solving problems that are too complex for conventional technologies — problems that do not have an algorithmic solution or for which an algorithmic solution is too complex to be found. In general, because of their abstraction from the biological brain, ANNs are well suited to problems that people are good at solving, but for which computers are not. These problems include pattern recognition and forecasting (which requires the recognition of trends in data).
ANNs are able to solve difficult problems in a way that resembles human intelligence. What is unique about neural networks is their ability to learn by example. Traditional artificial intelligence (AI) solutions rely on symbolic processing of the data, an approach which requires a priori human knowledge about the problem. Also neural networks techniques have an advantage over statistical methods of data classification because they are distribution-free and require no a priori knowledge about the statistical distributions of the classes in the data sources in order to classify them. Unlike these two approaches, ANNs are able to solve problems without any a priori assumptions. As long as enough data is available, a neural network will extract any regularity and form a solution.
An ideal neuron is assumed to respond optimally to the applied inputs. Neural network is a collective set of such neural units, in which the individual neurons are connected through complex synaptic connections characterized by weight coefficients. Every single neuron makes its contribution towards the computational properties of the whole system. The figure 5-1, shows a basic neuron.
The neuron has several input lines and a single output. Each input signal is weighted; that is, it is multiplied with the weight value of the corresponding input line (by analogy to the synaptic strength of the connections of real neurons). The neuron will combine these weighted inputs by forming their sum and, with reference to a threshold value and activation function; it will determine its output. In mathematical terms, the neuron is described by writing the following pair of equations:
u=∑_(i=1)^N▒〖w_i x_i 〗 ..(3.9)
and y = f (u -θ ) ..(3.10)
Where x1 , x2 , ….., xN are the input signals, w1 , w2 , ….., wN are the synaptic weights, u is the activation potential of the neuron, Ѳ is the threshold, y is the output signal of the neuron, and f (.) is the activation function.
The activation function, denoted by (.), defines the output of the neuron in terms of the activity level at its input. The most common form of activation function used in the construction of ANNs is the sigmoid function. An example of the sigmoid is the logistic function, defined by
f(u)=1/(1+exp(-au) ) ..(3.11)
Where is the slope parameter of the sigmoid function. By varying the parameter , we can obtain sigmoid functions of different slopes. In the limit, as the slope parameter approaches infinity, the sigmoid function becomes simply a threshold function. The threshold function, however, can take only the values 0 or 1, whereas a sigmoid function assumes a continuous range of values from 0 to 1. Also the sigmoid function is differentiable, whereas the threshold function is not. Differentiability is an important feature of neural network theory since it has a fundamental role in the learning process in ANNs.
Neural networks have emerged as advanced data mining tools in cases where other techniques may not produce satisfactory predictive models. Neural networks were inspired by models of biological neural networks since much of the motivation came from the desire to produce artificial systems capable of sophisticated, perhaps ‘intelligent’, computations similar to those that the human brain routinely performs, and thereby possibly to enhance our understanding of the human brain. Using three input variables-color, appearance and type of diseases were generated for calculating, training and evaluation of artificial neural networks.
The artificial neural network was designed with three neurons in the input layer (color, appearance, type of diseases) and one neuron in the output layer The optimum number of neurons in the hidden layer was obtained by using a trial and error method. Generally, networks with more hidden layers, less neurons and fewer width networks, have a better performance compared with networks of less depth and more neurons in one layer, although training in networks of less width is more difficult than for less depth networks.
An increasing method for selecting the number of neurons and layers was used for network training. New neurons were added to the network gradually, whenever in one stage network a local minimum was involved. This method has better practical potential for finding the correct size of a network. Advantages of this method are:
- Network entanglements were increased gradually with increasing neuron numbers,
- Optimum size of network was often obtained with this arrangement, and Monitoring and evaluation of local minimums were performed during training.
Actual algorithm for a 3-layer network (only one hidden layer):
Initialize the weights in the network (often randomly)
For each example e in the training set
- O = neural-net-output (network, e) ; forward pass
- T = teacher output for e
- Calculate error (T – O) at the output units
- Compute delta_wh for all weights from hidden layer to output layer ; backward pass
- Compute delta_wi for all weights from input layer to hidden layer;
- Backward pass continued
- Update the weights in the network
- Return the network
FEED– FORWARD BACK PROPAGATION NEURAL NETWORK
Back propagation is a systematic method for training multilayer artificial neural network. It has a mathematical foundation that is strong. It is a multilayer forward network using delta learning rule known as back propagation rule..Back propagation provides an efficient method for changing weights in a feed forward network, with differentiable activation function unit, to learn a training set of input- output examples.
Image classification is the final step in any pattern recognition problem. It is of two types. They are:
- Supervised classification and
- Unsupervised classification
In supervised classification, a priori knowledge of the images to be classified is known. Hence, the classification is simply a process of testing whether the computed classification agrees with the a priori knowledge. In unsupervised learning, there is not any a priori knowledge on the images to be classified. Hence, the classification is a little bit more tedious since we have no prior knowledge of the various data classes involved. There are various classification techniques. In this research, two classification approaches based on the supervised classification approach are implemented. They are listed below.
- Statistical classifier using the squared Mahalanobis minimum distance
- Neural network classifiers
- Multi layer feed-forward neural network with back propagation
- Radial basis function neural network
The word ‘Neural Network’ has been motivated from its inception by the recognition that the human brain computes in an entirely different way from the conventional digital computer. The brain is a highly complex, non-linear and parallel computer (information processing system). An artificial neural network (ANN), often just called a ‘neural network’ (NN), is a mathematical model or computational model based on biological neural network. It consists of an interconnected group of artificial network and processes information using a connectionist approach to computation. In most cases an ANN is an adaptive system that changes its structure based on external or internal information that flows through the network during the learning phase.
We have used a multilayered back propagation neural network (BPNN) as a classifier of different produce and in automatic detection of disease. The number of neurons in the input layer corresponds to the number of input features and the number of neurons in the output layer corresponds to the number of classes. The classifier is trained, validated and tested using images of different agriculture/horticulture produce.
The procedure adopted in classification is given in Algorithm.
Algorithm: BPNN Classifier.
- Step 1: Accept images of the agriculture/horticulture produce.
- Step 2: Extract different color and texture features.
- Step 3: Train the BPNN with extracted features.
- Step 4: Accept test images and perform.
- Step 5: Recognize and classify the produce images using BPNN classifier.
The basic structure of the BPNN includes one input layer, at least one hidden layer (single layer / multiple layers), followed by output layer. Neural network works by adjusting the weight values during training in order to reduce the error between the actual and desire output pattern. The aim of this network is to train the net to achieve a balance between the ability to respond correctly to the input patterns that are used for training and the ability to provide good responses to the input that are similar. The number of the node in the hidden layer is set by trail and error during training. The number of node in the output layer is equal to the number of the subjects in the database.
As we are giving 10 features to neural, its performance will be more accurate. The performance plot shows it very well. For the training here we have used 50 hidden layers neurons and 90000 epoch which gives better results. And to every disease a specific value is given which is the output of neural. Therefore, 10 inputs are given to neural and 1 desired output is given. Now when a new diseased image is given to system it extracts 10 features and then neural matches it with trained images and gives exact disease.
As from fig. 3.16 the performance plot of MSE and epochs is not much linear with 10 iterations as compared to plot in fig. 3.19 which is having 90000 iterations. For neural to be more accurate its performance should be linear, therefore by increasing the number of iterations the performance can be managed.