Structural Invariant Feature Segmentation Based Apple Fruit Disease Detection Using Deep Spectral Generative Adversarial Networks

In Indian economy agriculture, Crop cultivation plays an important role in the fruit production agricultural sector. At present, crop loss is primarily due to infested crops, resulting in reduced production rates for affecting various diseases. Manual monitoring of the disease is very difficult to analyze the type. It requires a huge amount of work, expertise, and excessive processing time.To tackle this problem, we introduce Deep Spectral Generative Adversarial Networks (DSGANs) algorithm to categorize the apple disease. Initially, the preprocessing was carried out through Median and Gabor Filters to enhance the frequency of the Image. Then the boundary regions are adaptively filtered with a canny edge detector. This supports the exact boundary regions of the object to get the affected region. The color variance and the contours are different from the affected and non-effected regions. To optimize this, the Self-Adaptive Plateau Histogram Equalization (SAPHE) technique
is applied to find the difference between affected and non-affected regions. Modified Gabor kernels are applied to choose the invariance of the affected region which supports segmentation using Invariant Sliding Window Segmentation (ISWS). This makes optimized segmentation by extracting the features in the affected boundary region. By intent, a ReLu activation for Logical decision to activate the logical decision depends on max successive threshold weights from convoluted margins. The non-linearity substitutive feature margins are extracted to enhance the performance of the output to process the adaptive GAN output layers. Finally, the classification part uses the DSGANs. The gated features get the marginal threshold values based on the feature extraction weights to get trained into classifier layers. This iteratively verifies the feature margins on affected regions get trained in Deep SpectralGenerative Adversarial Networks (DSGANs).