Automatic detection and removal of weeds is a challenging task that requires precise sensors. While crops and weeds possess similar features in terms of appearance, they can be discriminated based on spectral information. This can be done because any object has its own specific spectral signature based on its physical structure and chemical contents. This study examined the use of wavelet transform and deep learning for discrimination of weeds from crops. A total of 626 spectral reflectances in the range of 380 to 1,000 nm were obtained for three crops (cucumber [Cucumis sativus L.], tomato [Solanum lycopersicum L.], and bell pepper [Capsicum annuum L.]) and five different weeds (bindweed [Convolvulus spp.], purple nutsedge [Cyperus rotundus L.], narrowleaf plantain [Plantago lanceolata L.], common cinquefoil [Potentilla simplex Michx.], and garden sorrel [Rumex acetosa L.]). Morse wavelet was employed to decompose the spectra and extract the scalograms, which are the RGB representations of the spectral data. Two deep convolutional neural networks (i.e., GoogLeNet and SqueezNet) were employed for the recognition of crops and weeds. In addition, six common classifiers, including linear discriminant analysis, quadratic discriminant analysis, linear support vector machine, quadratic support vector machine, artificial neural networks, and k-nearest neighbors classifier (KNN), were used for the task of crop/weed discrimination to build the comparison with the proposed method. The error of prediction gradually decreased, and a 100% correct classification was achieved after 258 iterations. Analysis showed that SqueezNet provided classification of 100% accuracy, while GoogLeNet’s accuracy was 97.8% for the test set. Among the common classifiers, KNN provided the highest accuracy (i.e., 100%). This study showed that the proposed method can be successfully utilized for classification of crops and weeds.

Methods were developed and evaluated that utilize state of the art weed-sensing technology in row-crop production systems. Spectral differences in green living plants and bare soil allowed ‘real-time’ weed detection, with intermittent spraying of herbicide only where weeds were present. Sensor units were mounted in 0.7-m-wide hooded sprayers providing sensors with an unobstructed view of the area between soybean rows. Single hood and commercial-size eight-row systems were evaluated, and savings in glyphosate spray solution applied using sensors ranged from 63 to 85%, compared to conventional hooded spray systems with continuous application. Weed control by the sensor-controlled spray system was equal to the conventional system. This technology can significantly reduce herbicide usage and decrease production cost without reducing weed control.

Optimising oilseed rape canopy size through correct management is crucial for maximising yield. Plant growth regulators (PGRs) and nitrogen (N) fertiliser are generally applied at a flat rate, however variable applications may be useful for the optimisation of canopy size. The aim of this paper was to understand the potential for spectral reflectance indices to predict green area index (GAI) and crop N content in winter oilseed rape, with specific focus on the Fritzmeier Isaria Crop Sensor. Three large oilseed rape chessboard experiments were set up in 2015 and 2016 in the UK. The results show good correlations between the Isaria indices and both GAI and crop N content, suggesting that the Isaria may be a useful tool for variably applying PGRs and N fertiliser to oilseed rape.

A range of precision farming technologies are used commercially for variable rate applications of nitrogen (N) for cereals, yet these usually adjust N rates from a pre-set value, rather than predicting economically optimal N requirements on an absolute basis. This paper reports chessboard experiments set up to examine variation in N requirements, and to develop and test systems for its prediction, and to assess its predictability. Results showed very substantial variability in fertiliser N requirements within fields, typically >150 kg ha−1, and large variation in optimal yields, typically >2 t ha−1. Despite this, calculated increases in yield and gross margin with N requirements perfectly matched across fields were surprisingly modest (compared to the uniform average rate). Implications are discussed, including the causes of the large remaining variation in grain yield, after N limitations were removed.

Automated detection and identification of weeds in crop fields is the greatest obstacle to development of practical site-specific weed management systems. Research progress is summarized for two different approaches to the problem, remote sensing weed mapping and ground-based detection using digital cameras or nonimaging sensors. The general spectral and spatial limitations reported for each type of weed identification system are reviewed. Airborne remote sensing has been successful for detection of distinct weed patches when the patches are dense and uniform and have unique spectral characteristics. Identification of weeds is hampered by spectral mixing in the relatively large pixels (typically larger than 1 by 1 m) and will not be possible from imagery where weed seedlings are sparsely distributed among crop plants. The use of multispectral imaging sensors such as color digital cameras on a ground-based mobile platform shows more promise for weed identification in field crops. Spectral features plus spatial features such as leaf shape and texture and plant organization may be extracted from these images. However, there is a need for research in areas such as artificial lighting, spectral band requirements, image processing, multiple spatial resolution systems, and multiperspective images.

Human exploration missions to the Moon or Mars might be helped by the development of a bioregenerative advanced life-support (ALS) system that utilizes higher plants to regenerate water, oxygen and food. In order to make bioregenerative ALS systems competitive to physiochemical life-support systems, the ‘equivalent system mass’ (ESM) must be reduced by as much as possible. One method to reduce the ESM of a bioregenerative ALS system would be to deploy an automated remote sensing system within plant production modules to monitor crop productivity and disease outbreaks. The current study investigated the effects of canopy structure and imaging geometries on the efficiency of measuring the spectral reflectance of individual plants and crop canopies in a simulated ALS system. Results indicate that canopy structure, shading artefacts and imaging geometries are likely to create unique challenges in developing an automated remote sensing system for ALS modules. The cramped quarters within ALS plant growth units will create problems in collecting spectral reflectance measurements from the nadir position (i.e. directly above plant canopies) and, thus, crop canopies likely will be imaged from a diversity of orientations relative to the primary illumination source. In general, highly reflective white or polished surfaces will be used within an ALS plant growth module to maximize the stray light that is reflected onto plant canopies. Initial work suggested that these highly reflective surfaces might interfere with the collection of spectral reflectance measurements of plants, but the use of simple remote sensing algorithms such as 760/685 band ratios or normalized difference vegetation index (NDVI) images greatly reduced the effects of the reflective backgrounds. A direct comparison of 760/685 and NDVI images from canopies of lettuce, pepper and tomato plants indicated that unique models of individual plants are going to be required to properly assess the health conditions of canopies. A mixed model of all three plant species was not effective in predicting plant stress using either the 760/685 or NDVI remote sensing algorithms.