Automatic Localization and Count of Agricultural Crop Pests Based on an Improved Deep Learning Pipeline

Abstract

Insect pests are known to be a major cause of damage to agricultural crops. This paper proposed a deep learning-based pipeline for localization and counting of agricultural pests in images by self-learning saliency feature maps. Our method integrates a convolutional neural network (CNN) of ZF (Zeiler and Fergus model) and a region proposal network (RPN) with Non-Maximum Suppression (NMS) to remove overlapping detections. First, the convolutional layers in ZF Net, without average pooling layer and fc layers, were used to compute feature maps of images, which can better retain the original pixel information through smaller convolution kernels. Then, several critical parameters of the method were optimized, including the output size, score threshold, NMS threshold, and so on. To demonstrate the practical applications of our method, different feature extraction networks were explored, including AlexNet, ResNet and ZF Net. Finally, the model trained on smaller multi-scale images was tested on original large images. Experimental results showed that our method achieved a precision of 0.93 with a miss rate of 0.10. Moreover, our model achieved a mean Accuracy Precision (mAP) of 0.885.

Detecting Soil Microarthropods with a Camera-Supported Trap

“…Besides these "classical" techniques, passive acoustics was also used in monitoring studies [25]. While opto-electronic sensors have low power consumption and are less species-specific, cameras are just the opposite: They have a high energy demand and improved image analysis techniques for species determination [26]. There are other unique methods, such as the time-sorting pitfall trap, which records insects only for a limited time frame (e.g., days or dayparts) [16].…”

Putting deep learning in perspective for pest management scientists

“…Simple classification model to distinguish among some fungal infections of red chili peppers. Adapted from the expert system described in reference .…”

“…The process of deciding which images should be included in the training pool or in the test set against which the final model is evaluated can have major effects on training outcomes. Selection may be semi‐automated but nearly always includes some manual elements, either on the part of the person compiling the data set or indirectly – often before the image is posted on the internet . In addition, images are preprocessed to some degree before being submitted to the CNN, either by applying automated tools or manual selection to pick out objects (weeds or insects of interest) in training pool images …”

“…Much of the DL literature is steeped in jargon, but there are many excellent general reviews available that range from the very general to ones that focus on pharmaceutical or agricultural research and development. The examples of pest management applications discussed herein were chosen because they include good summaries of previous work in their respective areas.…”

Improved Winter Wheat Spatial Distribution Extraction Using A Convolutional Neural Network and Partly Connected Conditional Random Field

“…As these methods do not consider or barely consider the spatial correlation between pixels, the distinguishing ability of the extracted features is not ideal for several types of new higher-spatial-resolution remote sensing images.With the success of convolutional neural networks (CNNs) in camera image processing, researchers began to successfully use these networks for feature extraction from remote sensing images and have achieved good results [42][43][44][45]. The convolution operation can accurately express the spatial relationship between pixels and extract deep information from the pixels (when the convolution kernel is set appropriately), combining the advantages of previous feature extraction methods [14,[46][47][48]. Classic CNNs, such as the Fully Convolutional Network (FCN) [49], SegNet [50], DeepLab [51], and…”